sensors Article
A Wearable Wireless Sensor Network for Indoor Smart Environment Monitoring in Safety Applications Diego Antolín *, Nicolás Medrano, Belén Calvo and Francisco Pérez Grupo de Diseño Electrónico (I3A), Departamento de Ingeniería Electrónica y Comunicaciones, Universidad de Zaragoza, C/Pedro Cerbuna 12, Zaragoza 50009, Spain; [email protected] (N.M.); [email protected] (B.C.); [email protected] (F.P.) * Correspondence: [email protected]; Tel.: +34-876-55-34-27 Academic Editor: Gonzalo Pajares Martinsanz Received: 22 November 2016; Accepted: 8 February 2017; Published: 14 February 2017
Abstract: This paper presents the implementation of a wearable wireless sensor network aimed at monitoring harmful gases in industrial environments. The proposed solution is based on a customized wearable sensor node using a low-power low-rate wireless personal area network (LR-WPAN) communications protocol, which as a first approach measures CO2 concentration, and employs different low power strategies for appropriate energy handling which is essential to achieving long battery life. These wearables nodes are connected to a deployed static network and a web-based application allows data storage, remote control and monitoring of the complete network. Therefore, a complete and versatile remote web application with a locally implemented decision-making system is accomplished, which allows early detection of hazardous situations for exposed workers. Keywords: wearable sensor node; wearable wireless sensor network (W-WSN); CO2 smart detection; safety applications; remote web application
1. Introduction The growing advances in the last two decades in low-power wireless communications, the constant downsizing in electronic devices and the progressive increase of computational power in low-cost microcontrollers have fostered the emergence of cyber-physical systems (CPSs). A CPS [1,2] consists of a physical structure (a natural area, house, facility or even a human body) where a set of sensors monitors the main system parameters in order to obtain maximum knowledge and/or take advantage, with a minimum impact to the environment. Sensors interact with each other in order to obtain high-level information of the monitored physical system, to control some internal processes or to improve some characteristics, without the requirement of interaction with a central control system [3–8]. In safety monitoring, in industrial environments the usual approach lies in deploying static wireless sensor nodes in the area of interest. In this way, ambient monitoring systems based on wireless sensor networks (WSN) can be found in different scenarios such as refrigerated chambers [9], chemical production plants [10], or in modern steel mills to detect carbon monoxide [11]. In such potentially dangerous scenarios, continuous monitoring using wearable body sensor networks—added to work wear—will increase early detection of threatening situations for exposed workers. Wearable wireless sensor network (W-WSN) systems are a particular CPS case where sensors are deployed on the user clothing and/or body to monitor physiological parameters, environmental conditions, or both. Unlike conventional WSNs, W-WSNs consist of fewer and smaller nodes, covering less space. A typical wearable wireless sensor network consists of several wearable nodes connected between them or
Sensors 2017, 17, 365; doi:10.3390/s17020365
www.mdpi.com/journal/sensors
Sensors 2017, 17, 365
2 of 14
connected to a static WSN. Each wearable node includes low-power sensors, a wireless transceiver, electronic processing elements (microcontroller and interface systems) and the power supply unit, which must be miniaturized, lightweight and long lasting. This paper presents a complete W-WSN system based on a custom wearable sensor node using a low-power low-rate wireless personal area network LR-WPAN communications protocol, intended for monitoring ambient parameters in potentially harsh environments where carbon dioxide (CO2 ) gas leaks may occur. The proposed node contains two main sensors: CO2 concentration is measured by an IRC-A1 gas sensor from Alphasense (Essex, UK) with a measurement range from 0 to 50,000 ppm [12]; temperature and humidity are complementary parameters, measured by a SHT11 digital humidity sensor (Sensirion AG, Zurich, Switzerland) [13]. These nodes are designed to be connected to a deployed static network. A web-based application allows data storage, remote control and monitoring of the complete network, therefore achieving a complete and versatile remote web application with a locally-implemented decision-making system. The paper is structured as follows: Section 2 provides a brief background to better situate the present work. Section 3 presents the description of the sensor nodes, including the core hardware, microcontroller programing technologies and networking. Within the implementation strategies, special attention has been paid to reach optimal energy handling, essential to achieving long battery life. Section 4 comprises experimental results. Section 5 explains the operation of the complete remote web application system and finally, Section 6 presents the conclusions and future work. 2. Previous Work The proposed wearable device is based on a preliminary work [14], which served as the starting point to implement this optimized device that includes several both hardware and software improvements: an integrated battery charge system, static voltage scaling (SVS) and dynamic frequency scaling (DFS) techniques, and software routines that allow automatic node connection to a previously-deployed WSN. The CO2 sensor board includes a custom low-power astable circuit that excites the gas sensor, so the sensor response time keeps constant and independent of the microcontroller power mode. In addition, an acoustic buzzer has been included to advise the user about hazardous CO2 levels in the nearby environment. The wireless sensor network used as a fixed infrastructure where portable sensors can be added is based on a work by Antolín et al. [15]. When deployed indoors, the power consumption requirements of this static WSN can be relaxed by using suitable medium-sized batteries or even a mains power connection, so the duty cycle and number of physical external parameters to be measured per node can be increased without jeopardizing the system operating life. Static nodes are configured as router devices, while wearable sensors are configured as end devices. This guarantees that wearable nodes are always connected to the network through static devices, which allows its physical location to allow early detection of gas leakage. 3. Wearable Wireless Node Architecture 3.1. Wearable Sensor Node Hardware Design To achieve a truly wearable sensor device, key hardware design requirements are minimum size, efficient power management and a compact rechargeable battery system. Figure 1 shows the block diagram of the developed sensor node, and Figure 2 shows its photograph. It consists of a PIC18F26J50 8-bit microcontroller (µC) (Microchip Technology Inc., Chandler, AZ, USA), with nanoWatt technology, selected for its cost-characteristics tradeoff and low power consumption. It owns several energy-saving working modes and independent clock lines that can be addressed to different peripherals according to its timing requirements. It can be dynamically powered at a voltage ranging from 2.0 V to 3.6 V, with a quiescent current of 6.2 µA in non-sleep modes.
Sensors 2017, 17, 365
3 of 14
Sensors 2017, 17, 365 Sensors 2017, 17, 365
3 of 14 3 of 14
Figure 1. 1. Wearable sensor node block block diagram. Figure Wearable sensor Figure 1. Wearable sensor node node block diagram. diagram.
Figure 2. Wearable sensor node photograph: (Left) without XBee; (Right) with XBee. Figure 2. Wearable sensor node photograph: (Left) without XBee; (Right) with XBee. Figure 2. Wearable sensor node photograph: (Left) without XBee; (Right) with XBee.
The microcontroller manages the node operation, i.e., sensor acquisition; data collection and The microcontroller manages the node operation, i.e., sensor acquisition; data collection and storage; building of the data frame to benode sent by the radio frequency (RF) transceiver; sending of the Thebuilding microcontroller manages thebe i.e., sensor acquisition; data collection storage; of the data frame to sent operation, by the radio frequency (RF) transceiver; sending ofand the data frame to the of RFthe transceiver through an RS-232 protocol; and monitoring of the node state storage; building data frame to be sent by the radio frequency (RF) transceiver; sending data frame to the RF transceiver through an RS-232 protocol; and monitoring of the node state (energy mode, initialization, etc.). The node RF transceiver is an XBee module with a wire antenna to of the data frame to the RF etc.). transceiver through an RS-232 protocol; and monitoring the node (energy mode, initialization, The node RF transceiver is an XBee module with a wireofantenna to reduce the size. It includes the DM-24 firmware from DigiMesh that works in the 2.4 GHz state (energy mode, initialization, etc.). The node RF transceiver is an XBee module with wire reduce the size. It includes the DM-24 firmware from DigiMesh that works in the 2.4a GHz Industrial-Scientific-Medical (ISM) band with an IEEE 802.15.4 protocol running in the transceiver, antenna to reduce the size. It includes the with DM-24 from protocol DigiMeshrunning that works in the 2.4 GHz Industrial-Scientific-Medical (ISM) band anfirmware IEEE 802.15.4 in the transceiver, making compatible the connection of this with device toIEEE a deployed WSN and running allowingin drawing up of its Industrial-Scientific-Medical (ISM) band an 802.15.4 protocol the transceiver, making compatible the connection of this device to a deployed WSN and allowing drawing up of its communications structure. making compatible the connection of this device to a deployed WSN and allowing drawing up of its communications structure. The sensor node includes a novel and compact power supply system similar to those available communications structure. The sensor node includes a novel and compact power supply system similar to those available in smartphones and other portable consumer electronic devices. It consists of a MCP73833 battery The sensor node includes a novel and compact powerdevices. supply system similar those available in in smartphones and other portable consumer electronic It consists of atoMCP73833 battery charger circuit from Microchip, a rechargeable battery and two low dropout (LDO) regulators to smartphones and other portable consumer electronic devices. consists a MCP73833 charger charger circuit from Microchip, a rechargeable battery andIttwo low of dropout (LDO)battery regulators to provide the power voltage levels required by the different components. The selected battery is a 3.7 V circuit Microchip, a rechargeable battery and two lowcomponents. dropout (LDO) to provide providefrom the power voltage levels required by the different The regulators selected battery is a 3.7the V and 800 mAh Lithium Polymer (LiPo) LP-573442-1S-3 that meets the W-WSN node energy power voltage required by the (LiPo) differentLP-573442-1S-3 components. The selected is a 3.7 V node and 800 mAh and 800 mAhlevels Lithium Polymer that meetsbattery the W-WSN energy requirements with a small size. The voltage regulators provide stable, well-defined voltage levels Lithium Polymer (LiPo) LP-573442-1S-3 that meets the W-WSN node energy requirements with a levels small requirements with a small size. The voltage regulators provide stable, well-defined voltage from The the voltage decreasing batteryprovide voltage:stable, an MCP1725 LDO voltage voltage levels regulator (Microchip Technology size. regulators well-defined from the decreasing battery from the decreasing battery voltage: an MCP1725 LDO voltage regulator (Microchip Technology Inc., Chandler, AZ, USA) provides a 3 V constant voltage to power the XBee RF transceiver. The voltage: an MCP1725 LDO voltage regulator (Microchip Technology Inc.,the Chandler, AZ, USA) provides Inc., Chandler, AZ, USA) provides a 3 V constant voltage to power XBee RF transceiver. The second LDO regulator (TC1015 also from Microchip) provides a constant 2.5 V level to power the asecond 3 V constant voltage to(TC1015 power the XBee RFMicrochip) transceiver.provides The second LDO regulator (TC1015 also from LDO regulator also from a constant 2.5 V level to power the rest of node electronics. Microchip) a constant 2.5 V level to power the rest of node electronics. rest of nodeprovides electronics. Sensors are housed on a specific printed printed circuit board board (Figure 3) 3) connected to to the sensor sensor node Sensors are housed Sensors are housed on on aa specific specific printed circuit circuit board (Figure (Figure 3) connected connected to the the sensor node node through a pin in-line connector. This allows an easy replacement for damaged or obsolete sensors through through aa pin pin in-line in-line connector. connector. This This allows allows an an easy easy replacement replacement for for damaged damaged or or obsolete obsolete sensors sensors without modifying the the main node node architecture. architecture. The version herein presented measures of humidity, without The version version herein herein presented presented measures measures of of humidity, humidity, without modifying modifying the main main node architecture. The temperature and CO 2 concentration. A SHT11 (Sensirion AG, Zurich, Switzerland) provides the temperature A SHT11 SHT11 (Sensirion (Sensirion AG, AG, Zurich, Zurich, Switzerland) temperature and and CO CO22 concentration. concentration. A Switzerland) provides provides the the relative humidity humidityand andtemperature temperatureinformation, information,with withan anoperation operationrange rangeofof0% 0%toto100% 100%and and−−40 °C to relative 40 ◦°C C to relative humidity and temperature information, with an operation range of 0% to 100% and −40 to 124 °C, respectively. The gas detection is performed by a non-dispersive infrared IRC-A1 CO 2 sensor ◦ C, respectively. The gas detection is performed by a non-dispersive infrared IRC-A1 CO sensor 124 124 °C, respectively. The gas detection is performed by a non-dispersive infrared IRC-A1 CO22 sensor from Alphasense (Essex, UK) [12,16] in the range of 0 to 50,000 ppm to match safety applications. from Alphasense (Essex, UK) [12,16] in the range of 0 to 50,000 ppm to match safety applications. This sensor requires 2 Hz square wave excitation, a preheating time of 30 min and up to 40 s of This sensor requires 2 Hz square wave excitation, a preheating time of 30 min and up to 40 s of
Sensors 2017, 17, 365
4 of 14
Sensors 2017, 17, 365
4 of 14
from Alphasense (Essex, UK) [12,16] in the range of 0 to 50,000 ppm to match safety applications. response time. Our conditioning circuitry has been adapted from [17], fitand the up µCtoanalog-to-digital This sensor requires 2 Hz square wave excitation, a preheating time of 30 to min 40 s of response converter (ADC) input voltage requirements. time. Our conditioning circuitry has been adapted from [17], to fit the µC analog-to-digital converter The 2-Hz squarerequirements. signal required by the sensor is directly provided by an independent astable (ADC) input voltage circuit based operational (Maxim Integrated, San José, USA) instead of a The 2-Hzonsquare signal amplifier required MAX4038 by the sensor is directly provided by anCA, independent astable digital output from the microcontroller. Thus, the microcontroller can be set to a low-power mode, circuit based on operational amplifier MAX4038 (Maxim Integrated, San José, CA, USA) instead of the energy while Thus, the sensor keeps always can active to minimize its response areducing digital output from consumption, the microcontroller. the microcontroller be set to a low-power mode, time. reducing the energy consumption, while the sensor keeps always active to minimize its response time.
(a)
(b) Figure 3. Sensor connection: (a) photograph. Figure 3. Sensor board board and and node node connection: (a) block block diagram; diagram; (b) (b) photograph.
The three signals (reference, active and temperature voltages) in the IRC-A1 sensor must be three signals (reference, active and temperature voltages) in the IRC-A1 sensor must be used usedThe to calculate the CO2 concentration, by applying the equations given in the sensor datasheet. to calculate the CO2 concentration, by applying the equations given in the sensor datasheet. However, However, due to the complexity of the required equations and the use of a low-cost microcontroller due the complexity of thethe required equations and the use of a low-cost microcontroller as a sensor as a to sensor node manager, microcontroller includes look-up table (LUT) data that allows for a node manager, the microcontroller includes look-up table (LUT) data that allows for a coarse estimate coarse estimate of the measured CO2 levels, while an acoustic alarm using a buzzer included in the of the measured CO while an acoustic using a buzzer included in the board is activated 2 levels, board is activated when concentration exceedsalarm a predetermined level. when concentration exceeds a predetermined level. 3.2. Wearable Sensor Node Software 3.2. Wearable Sensor Node Software Figure 4 shows the operation flowchart of the wearable sensor node software. Once the power Figure 4 shows the operation flowchart of the wearable sensor node software. Once the power is is switched on for the very first time, the microcontroller is firstly configured using locally-stored switched on for the very first time, the microcontroller is firstly configured using locally-stored default default parameters: duty/sleep timing, operation frequencies, etc. Next, the RF transceiver is parameters: duty/sleep timing, operation frequencies, etc. Next, the RF transceiver is initialized initialized following a microcontroller request, initializing the universal asynchronous following a microcontroller request, initializing the universal asynchronous receiver-transmitter receiver-transmitter (UART), configuring the interrupt service routine (ISR) and starting the wireless (UART), configuring the interrupt service routine (ISR) and starting the wireless network joint process. network joint process. Network joining requires an XBee software reset, then a commissioning Network joining requires an XBee software reset, then a commissioning command is sent to the network command is sent to the network coordinator node to indicate that a new device is joining the network, and finally the transceiver is sent to sleep mode. The delays shown in the flowchart are required by the transceiver for a suitable processing of the commands.
Sensors 2017, 17, 365
5 of 14
coordinator node to indicate that a new device is joining the network, and finally the transceiver is sent to sleep mode. The delays shown in the flowchart are required by the transceiver for a suitable processing of2017, the17,commands. Sensors 365 5 of 14 Once both microcontroller and transceiver have been initialized, sensor input/outputs and Once both microcontroller and transceiver have been initialized, sensor input/outputs and peripherals (such as the analog-to-digital converter) are configured. Configuration is completed by peripherals (such as the analog-to-digital converter) are configured. Configuration is completed by enabling the microcontroller to receive interrupts from the transceiver, then driving the node to sleep enabling the microcontroller to receive interrupts from the transceiver, then driving the node to mode. Once node operation of aconsists main loop, the microcontroller is sleep fully mode.configured, Once fully the configured, the nodeconsists operation of a where main loop, where the awakened by an interrupt request the XBee. Next, increases microcontroller is awakened by from an interrupt request fromthe the operation XBee. Next, frequency the operation frequencyin order increases in order to time, reducethe the sensor acquisition sensor data are collected, to the in the to reduce the acquisition datatime, are the collected, compared to thecompared values stored values the LUT coordinator and sent to thethrough networkthe coordinator through the transceiver. Finally, the LUT and sentstored to theinnetwork transceiver. Finally, the operation frequency is operation frequency is decreased and the system goes back to sleep mode until a new XBee interrupt decreased and the system goes back to sleep mode until a new XBee interrupt is produced. is produced.
Figure 4. Flowchart of sensor node software operation. ISR: Interrupt service routine
Figure 4. Flowchart of sensor node software operation. ISR: Interrupt service routine. 4. Network Test
4. NetworkThe Testproposed system has been tested in the indoor deployment shown in Figure 5. Each room includes a static node (green), which sends parameters throughshown multi-hop to room The proposed system has been tested inambient the indoor deployment in transmission Figure 5. Each the node coordinator (orange, room #3), and then information is sent to a host device. includes a static node (green), which sends ambient parameters through multi-hop transmission to the node coordinator (orange, room #3), and then information is sent to a host device.
Sensors 2017, 17, 365
6 of 14 6 of 14 6 of 14
Room #1 #1 Room
Room #4 #4 Room
Sensors 2017, 17, 365 Sensors 2017, 17, 365
Figure 5. Indoor wearable wireless sensor network (W-WSN) deployment in the application test Figure 5. Indoor wearable wireless sensor network (W-WSN) deployment in the application test Figure 5. (dimensions: Indoor wearable scenario 50 m wireless × 15 m). sensor network (W-WSN) deployment in the application test scenario (dimensions: 50 m × 15 m). scenario (dimensions: 50 m × 15 m).
Wearable nodes (red, Figure 5) connect to the deployed network through the nearest fixed Wearable nodes (red, 5)5)connect to the deployed network throughthe thenearest nearestfixed fixed Wearable nodes (red,Figure Figure connect tothe theowner deployed network through device, therefore providing information about position and displacement. Because data device, therefore providing information position and displacement. Because data device, therefore providing information aboutthe the owner owner position and displacement. Because data transmissions are performed over shortabout distances, RF transmission power in mobile devices is set to transmissions are performed over short distances, RF power in mobile devices is setistoset transmissions are performed over short distances, RFtransmission transmission power in mobile devices the minimum available value, i.e., 10 dBm for XBee modules [18]. Node locations are upgraded the minimum value, i.e., for XBee modules [18]. Node locations are upgraded to the minimum available value,send i.e., 10 10 dBm dBm forhost XBee modules [18]. Node locations areevery upgraded every time theavailable network nodes data to the system. Busy periods are configured 15 s, every time the network nodes send data to the host system. Busy periods are configured every 1515 s, s, every timeallthe nodes send data thefrom hostthe system. Busy periods are configured every so the thenetwork monitored parameters areto sent wearable and static nodes via the host system so the all the monitored parameters are sent from the wearable and static nodes via the host system to all thethe processing application which processes thewearable data, activating suitable protocols insystem case ofto so the monitored parameters are sent from the and static nodes via the host the processing application processes the activating data, activating suitable protocols of unusual magnitude levels. theto processing application whichwhich processes the data, suitable protocols in caseinofcase unusual unusual magnitude levels. To check the correct operation of the proposed W-WSN, wearable nodes were positioned on magnitude levels. To check the correct operation of the proposed W-WSN, wearable nodes were positioned on volunteers along the monitored Figure W-WSN, 6 shows the measurements provided by one of To check moving the correct operation of thearea. proposed wearable nodes were positioned on volunteers moving alonga the monitored area. Figure days, 6 shows the measurements provided bypresents one of the volunteers wearing node for three consecutive for periods of 3.5–4.5 h. Figure 6a volunteers moving along the monitored area. Figure 6 shows the measurements provided by one the volunteers wearing a node for threewhile consecutive for periods of 3.5–4.5 h. Figure 6a presents measurement for Figure days, 6b shows the periods CO 2 concentration. 2 peaks of the volunteers results wearing a humidity, node for three consecutive days, for of 3.5–4.5 h.COFigure 6a measurement results for humidity, while Figure 6b shows the CO 2 concentration. CO2 peaks correspond to measurements when the user entered one of the dependencies occupied 26 presents measurement results for humidity, while Figure 6b shows the CO2 concentration. CO2bypeaks correspond students. to measurements when the user entered one of the dependencies occupied by 26 correspond to measurements when the user entered one of the dependencies occupied by 26 students. students. To validate the network link quality, the different packet error ratios (PER) have been To To validate the network link quality, the different packet error ratios (PER) have beenhave determined. validate the network link quality, the static different (PER) been determined. The results (Table 1) show a PER for nodespacket higher error than inratios the case of mobile nodes Thedetermined. results (Table 1) show a PER for static nodes higher than in the case of mobile nodes connected to results (Table 1) show PER static than the caseofofsight, mobile nodes connected toThe static devices, mainly duea to thefor fact thatnodes static higher links are in in non-line while the static devices,tomainly due to the fact due thatto static links arestatic in non-line of sight, while mobile nodes connected static mainly the fact that links are non-line of the sight, while the mobile nodes are devices, connected to the network with line of sight. Theinincrease of errors in packet aremobile connected to the network with line of sight. The increase of errors in packet transmission due nodes due are to connected to the network withtoline of sight. Thelocation. increase of errors in packet to transmission node reconnections is related changes in user node reconnections is related to changes in location. transmission due to node reconnections is user related to changes in user location. 46.5
R elative humhum idityidity (% )(% ) R elative
46.5 46 46 45.5
45.5 45 45 44.5
44.5 44 44 43.5
43.5 43 43 42.5 0 30 60 90 120 150 180 210 30 60 90 120 150 180 210 240 270 30 60 90 120 150 180 210 240 270 42.5 (min)210 240 270 30 60 90 120 150 180 210 240 270 0 30 60 90 120 150 180 210 30 60 90 120 Time 150 180
Time (a)(min)
(a)
Figure 6. Cont. Figure 6. Cont. Figure 6. Cont.
Sensors 2017, 17, 365 Sensors 2017, 17, 365
7 of 14 7 of 14
750
Sensors 2017, 17, 365 CO2 concentration (ppm ) (ppm ) CO2 concentration
700 750 650 700 600 650 550 600 500 550 450 500 400 0 450 400 0
7 of 14
30
60
90
120 150 180 210 30
60
90 120 150 180 210 240 270 30
30
60
90
120 150 180 210 30
60
90 120 150 180 210 240 270 30
Time (min)
(b)
60
90
120 150 180 210 240 270
60
90
120 150 180 210 240 270
(min) Figure 6. (a) Relative humidity acquired duringTime laboratory sessions; (b) CO2 concentration acquired Figure 6. (a) Relative humidity acquired during laboratory sessions; (b) CO2 concentration acquired (b) during these same periods. during these same periods. Figure 6. (a) Relative humidity acquired during laboratory sessions; (b) CO2 concentration acquired Table 1. Packet Error Ratio. during these same periods. Table 1. Packet Error Ratio.
Node
Table 1. Packet Error Ratio. Static Node Node
Rate (%) Rate 2.6 (%)
Mobile Node Connected 0.95 Node Rate (%) Static Nodeto Static Router 2.6 Mobile Node Reconnecting to Other Router 3.4 Static Node 2.6 Mobile Node Connected to Static Router 0.95 Mobile to Static Other Router Router 3.4 MobileNode NodeReconnecting Connected to 0.95 An important issue in theNode design of a wearable sensor node3.4 is its operating lifetime. To Mobile Reconnecting to wireless Other Router obtain a reliable node lifetime estimation, the evolution of the battery voltage and energy level in An important issue in the design of a wearable wireless sensor node is its operating lifetime. real operation conditions measured and numerically modeled. was verified that the An important issue inhas thebeen design of a wearable wireless sensor nodeFirst, is itsitoperating lifetime. To To obtain a reliable node lifetime estimation, the evolution of the battery voltage and energy level experimental power consumption is constant in each one of the node states along time. Then, to obtain a reliable node lifetime estimation, the evolution of the battery voltage and energy level in in real operation conditions has been measured and numerically modeled. First, it was verified that obtain the battery discharge for the 3.7 V, 800 mAh LiPo LP-573442-1S-3 battery, the current node real operation conditions has been measured and numerically modeled. First, it was verified that the theconsumption experimentalprofile powerwas consumption is constant in eachvoltage one of values the node states7). along time. Then, measured for several battery (Figure As the figure experimental power consumption is constant in each one of the node states along time. Then, to to obtain the battery discharge for the 3.7 V, 800 mAh LiPo LP-573442-1S-3 battery, the current node shows,the thebattery currentdischarge consumption is also independent of the battery outputbattery, voltage:the 5.5current mA in node sleep obtain for the 3.7 V, 800 mAh LiPo LP-573442-1S-3 consumption profile measured forfor several battery voltage values (Figure 7). As the shows, mode (mainly due was to was keeping the CO 2 sensor powered to voltage allow fast measurement without for consumption profile measured several battery values (Figure 7). As figure theneed figure theshows, current consumption is also independent of the battery output voltage: 5.5 mA in sleep mode warm-up time), 69 mA in acquisition mode and 60 mA in RF transmission. This is due to the use of the current consumption is also independent of the battery output voltage: 5.5 mA in sleep (mainly due to keeping the CO sensor powered to allow fast measurement without need for warm-up LDO voltage regulators to power the node at fixed voltages from the battery varying voltage. 2 mode (mainly due to keeping the CO2 sensor powered to allow fast measurement without need for time), 69 mAtime), in acquisition and 60 mA and in RF60transmission. This is dueThis to the voltage warm-up 69 mA in mode acquisition mode mA in RF transmission. is use dueof toLDO the use of regulators to power the node at fixed fromvoltages the battery voltage. LDO voltage regulators to power thevoltages node at fixed fromvarying the battery varying voltage.
Figure 7. Mobile node current consumption. Nodes are biased by a nominal 3.7 V battery, charged to 5 V (green), 4.2 V (red) and 3.4 V (blue). Figure Mobilenode nodecurrent currentconsumption. consumption. Nodes 3.73.7 VV battery, charged to to Figure 7. 7. Mobile Nodesare arebiased biasedby bya anominal nominal battery, charged 5 V (green), 4.2 V (red) and 3.4 V (blue). 5 V (green), 4.2 V (red) and 3.4 V (blue).
Sensors 2017, 17, 365
8 of 14
The battery discharge at constant current is shown in Figure 8: the battery voltage drastically Sensors 2017, 17, 365 8 of 14 drops after its value falls below 3.5 V. For the useful region, it is modeled according to The battery discharge at constant current is shown in Figure 8: the battery voltage drastically
13 10 v(tdrops 1.08 10−falls × t5 +3.5 1.493 ×the 10− ×region, t4 − 7.36 × 10−8 × t3 + 1.729 × 10−5 × t2 )= − after its × value below V. For useful it is modeled according to
−3 − 3.047 ×= 10 v in volts, × +(1.493 × 10 t×minutes − 7.36 )× 10 −1.08× × t10+ 4.125 Sensors 2017, 17, 365
10
×
− 3.047 × 10
× + 1.729 × × + 4.125 (v in volts, t in minutes)
(1)
8 (1) of 14
The battery discharge at constant current is shown in Figure 8: the battery voltage drastically drops after its value falls below 3.5 V. For the useful region, it is modeled according to × + 1.493 × 10 × − 7.36 × 10 × + 1.729 × = −1.08 × 10 10 × − 3.047 × 10 × + 4.125 (v in volts, t in minutes)
(1)
Figure 8. Experimental battery voltage (V) vs. time (min) for a constant 60 mA discharge.
Figure 8. Experimental battery voltage (V) vs. time (min) for a constant 60 mA discharge. Based on the results presented in Figure 8, the relationship between the remaining energy and the voltage provided by the battery can be obtained. The result is presented in Figure 9. This Based on the results presented in Figure 8, the relationship between the remaining energy behavior is approximately given by:
and the voltage provided by the battery can be obtained. The result is presented in Figure 9. This behavior is Figure 8. Experimental battery voltage (V) vs. time (min) for a constant 60 mA discharge. approximately given by: Based on the results presented in Figure 8, the relationship between the remaining energy and the voltage provided 5by the2 battery can be 6obtained. The result is6 presented in Figure 9. This C (v) = −1.243 × 10 × v + 1.026 × 10 × v − 2.073 × 10 (v in volts, C in mAmin) behavior is approximately given by:
(2)
Figure 9. Experimental (blue) and polynomial (red) fit for battery discharge (mAmin) vs. battery voltage. = −1.243 × 10 ×
+ 1.026 × 10 ×
− 2.073 × 10 (v in volts, C in mAmin)
(2)
Figure Experimental (blue) and polynomial (red) fit for discharge battery discharge battery On the 9.other hand, the equation of the node battery in each(mAmin) workingvs.cycle can be Figure 9. Experimental (blue) and polynomial (red) fit for battery discharge (mAmin) vs. voltage. obtained according to the process described in [15] as follows
battery voltage.
= −1.243 × 10 ×
C=
t1
t2
t3
i1 (t ) dt i2 ( t ) dt K i3 ( t )dt
t0 + 1.026 × 10 ×t 1 − 2.073 × 10t 2 (v in volts, C in mAmin)
(3) (2)
On the other hand, the equation of of the batterydischarge discharge in working each working can be On the other hand, the equation thenode node battery in each cycle cancycle be obtained according to the process described inin[15] follows obtained according to the process described [15] as follows C=
t0
t1
t
t
2 3 i (t ) dtZ t2 i2 ( t ) dt K Zi3 (tt3)dt t1 t2 i1 (t)dt + i2 (t)dt + K i3 (t)dt t1 t2
Z tC= 1
t0 1
(3)
(3)
where ix (t) represents the current in the three different node states: sleep, with time limits t0 and t1 ; measure, with time limits t1 and t2 ; and RF transmission, with time limits t2 and t3 . Since the current consumption in each state (see Figure 7) is constant, Equation (3) can be simplified to
Sensors 2017, 17, 365 Sensors 2017, 17, 365
9 of 14
9 of 14
where ix(t) represents the current in the three different node states: sleep, with time limits t0 and t1; measure, with time limits t1 and t2; and RF transmission, with time limits t2 and t3. Since the current − t2be (4) − t1 ) + i3(3) C Figure = i1 (t17)−ist0constant, (t3 can ) simplified to ) + i2 (t2 Equation consumption in each state (see C=i 1 t1the t 0battery i2 t 2discharge t1 i3 t 3asat 2function In this way, it is possible to estimate of time as
In this way, it is possible to estimate the battery discharge as a function of time as C (t) = −6.575 × t + 47998.38 (C in mAmin, t in minutes) = −6.575 × + 47998.38 (C in mAmin, t in minutes)
(4)
(5) (5)
Taking into account operatingvoltage voltageisislimited limited V, which corresponds Taking into accountthat thatthe the useful useful operating to to 3.53.5 V, which corresponds to a to a remaining energy of approximately 8000 mAmin (Figure 9), simulation results indicate a wearable remaining energy of approximately 8000 mAmin (Figure 9), simulation results indicate a wearable node operating 5.34 days daysexperimentally experimentallydetermined, determined, node operatinglife lifeofof5.07 5.07days days(Figure (Figure10), 10), close close to to the the 5.34 forfor a a duty cycle ofof 73.1 valuesensure ensureaasuitable suitableoperating operating duty cycle 73.1ms msand and1515s sofofworking workingcycle. cycle. These These values lifelife forfor a a wearable device, setting down downananappropriate appropriate schedule for battery recharge processes, wearable device,allowing allowing setting schedule for battery recharge processes, i.e., i.e.,this this operability time is long enough a portable system be charged on a basis. daily The basis. operability time is long enough for afor portable system to be to charged on a daily rechargeable system integrated in mobile sensor node node is essential for thisfor application and it isand an it The rechargeable system integrated in mobile sensor is essential this application improvement with respect to previous work. is an improvement with respect to previous work. Finally, Table 2 summarizes the main performances the proposed wearable node and Finally, Table 2 summarizes the main performances of theof proposed wearable node and compares compares them with other similar solutions. them with other similar solutions.
Figure Lifetimeestimation. estimation. Red Red line line indicates indicates the inin the battery. Figure 10.10.Lifetime theuseful usefulenergy energylimit limitstored stored the battery. Table 2. Sensor Nodes Comparison. Table 2. Sensor Nodes Comparison. Device Parameters
IRIS Crossbow IRIS Crossbow
Micaz Crossbow Micaz Crossbow
TelosB Crossbow TelosB Crossbow MSP430 16 bits MSP430 N/A 16 bits 1.8 mA N/A 5.1mA µA 1.8
Waspmote Libelium Waspmote Libelium ATMega1281 8 bits ATMega1281 8 MHz 8 bits 9 mA 8 MHz 62 µA 9 mA
This Work
Device Parameters This Work Processor Microcontroller ATMega1281 ATMega1281 PIC18F26J50 Processor N° Bits 8 bits 8 bits 8 bit Microcontroller ATMega1281 ATMega1281 PIC18F26J50 Frequency N/A N/A 8 MHz ◦ N Bits 8 bits 8 bits 8 bit Active Mode Current 8 mA 8 mA 7 mA Frequency N/A N/A 8 MHz Sleep ModeCurrent Current 8 µA 50 m 20–30 m 20–30 m 60–90 m 60–90 m Max. Tx Power 3 dBm 0 dBm 0 dBm 18 dBm 18 dBm Sensitivity −101 dBm −94 dBm −94 dBm −100 dBm −100 dBm Receive Mode 16 mA 19.7 mA 23 mA 57.08 mA 56.4 mA Max. Tx Power 3 dBm 0 dBm 0 dBm 18 dBm 18 dBm Transmission Current mA 17.4 mA 188 mA 69 mA Receive Mode 1617 mA 19.7 mA 23 N/A mA 57.08 mA 56.4 mA Sleep Mode NA 1 µA 1 µA 120 µA
A Wearable Wireless Sensor Network for Indoor Smart Environment Monitoring in Safety Applications Diego Antolín *, Nicolás Medrano, Belén Calvo and Francisco Pérez Grupo de Diseño Electrónico (I3A), Departamento de Ingeniería Electrónica y Comunicaciones, Universidad de Zaragoza, C/Pedro Cerbuna 12, Zaragoza 50009, Spain; [email protected] (N.M.); [email protected] (B.C.); [email protected] (F.P.) * Correspondence: [email protected]; Tel.: +34-876-55-34-27 Academic Editor: Gonzalo Pajares Martinsanz Received: 22 November 2016; Accepted: 8 February 2017; Published: 14 February 2017
Abstract: This paper presents the implementation of a wearable wireless sensor network aimed at monitoring harmful gases in industrial environments. The proposed solution is based on a customized wearable sensor node using a low-power low-rate wireless personal area network (LR-WPAN) communications protocol, which as a first approach measures CO2 concentration, and employs different low power strategies for appropriate energy handling which is essential to achieving long battery life. These wearables nodes are connected to a deployed static network and a web-based application allows data storage, remote control and monitoring of the complete network. Therefore, a complete and versatile remote web application with a locally implemented decision-making system is accomplished, which allows early detection of hazardous situations for exposed workers. Keywords: wearable sensor node; wearable wireless sensor network (W-WSN); CO2 smart detection; safety applications; remote web application
1. Introduction The growing advances in the last two decades in low-power wireless communications, the constant downsizing in electronic devices and the progressive increase of computational power in low-cost microcontrollers have fostered the emergence of cyber-physical systems (CPSs). A CPS [1,2] consists of a physical structure (a natural area, house, facility or even a human body) where a set of sensors monitors the main system parameters in order to obtain maximum knowledge and/or take advantage, with a minimum impact to the environment. Sensors interact with each other in order to obtain high-level information of the monitored physical system, to control some internal processes or to improve some characteristics, without the requirement of interaction with a central control system [3–8]. In safety monitoring, in industrial environments the usual approach lies in deploying static wireless sensor nodes in the area of interest. In this way, ambient monitoring systems based on wireless sensor networks (WSN) can be found in different scenarios such as refrigerated chambers [9], chemical production plants [10], or in modern steel mills to detect carbon monoxide [11]. In such potentially dangerous scenarios, continuous monitoring using wearable body sensor networks—added to work wear—will increase early detection of threatening situations for exposed workers. Wearable wireless sensor network (W-WSN) systems are a particular CPS case where sensors are deployed on the user clothing and/or body to monitor physiological parameters, environmental conditions, or both. Unlike conventional WSNs, W-WSNs consist of fewer and smaller nodes, covering less space. A typical wearable wireless sensor network consists of several wearable nodes connected between them or
Sensors 2017, 17, 365; doi:10.3390/s17020365
www.mdpi.com/journal/sensors
Sensors 2017, 17, 365
2 of 14
connected to a static WSN. Each wearable node includes low-power sensors, a wireless transceiver, electronic processing elements (microcontroller and interface systems) and the power supply unit, which must be miniaturized, lightweight and long lasting. This paper presents a complete W-WSN system based on a custom wearable sensor node using a low-power low-rate wireless personal area network LR-WPAN communications protocol, intended for monitoring ambient parameters in potentially harsh environments where carbon dioxide (CO2 ) gas leaks may occur. The proposed node contains two main sensors: CO2 concentration is measured by an IRC-A1 gas sensor from Alphasense (Essex, UK) with a measurement range from 0 to 50,000 ppm [12]; temperature and humidity are complementary parameters, measured by a SHT11 digital humidity sensor (Sensirion AG, Zurich, Switzerland) [13]. These nodes are designed to be connected to a deployed static network. A web-based application allows data storage, remote control and monitoring of the complete network, therefore achieving a complete and versatile remote web application with a locally-implemented decision-making system. The paper is structured as follows: Section 2 provides a brief background to better situate the present work. Section 3 presents the description of the sensor nodes, including the core hardware, microcontroller programing technologies and networking. Within the implementation strategies, special attention has been paid to reach optimal energy handling, essential to achieving long battery life. Section 4 comprises experimental results. Section 5 explains the operation of the complete remote web application system and finally, Section 6 presents the conclusions and future work. 2. Previous Work The proposed wearable device is based on a preliminary work [14], which served as the starting point to implement this optimized device that includes several both hardware and software improvements: an integrated battery charge system, static voltage scaling (SVS) and dynamic frequency scaling (DFS) techniques, and software routines that allow automatic node connection to a previously-deployed WSN. The CO2 sensor board includes a custom low-power astable circuit that excites the gas sensor, so the sensor response time keeps constant and independent of the microcontroller power mode. In addition, an acoustic buzzer has been included to advise the user about hazardous CO2 levels in the nearby environment. The wireless sensor network used as a fixed infrastructure where portable sensors can be added is based on a work by Antolín et al. [15]. When deployed indoors, the power consumption requirements of this static WSN can be relaxed by using suitable medium-sized batteries or even a mains power connection, so the duty cycle and number of physical external parameters to be measured per node can be increased without jeopardizing the system operating life. Static nodes are configured as router devices, while wearable sensors are configured as end devices. This guarantees that wearable nodes are always connected to the network through static devices, which allows its physical location to allow early detection of gas leakage. 3. Wearable Wireless Node Architecture 3.1. Wearable Sensor Node Hardware Design To achieve a truly wearable sensor device, key hardware design requirements are minimum size, efficient power management and a compact rechargeable battery system. Figure 1 shows the block diagram of the developed sensor node, and Figure 2 shows its photograph. It consists of a PIC18F26J50 8-bit microcontroller (µC) (Microchip Technology Inc., Chandler, AZ, USA), with nanoWatt technology, selected for its cost-characteristics tradeoff and low power consumption. It owns several energy-saving working modes and independent clock lines that can be addressed to different peripherals according to its timing requirements. It can be dynamically powered at a voltage ranging from 2.0 V to 3.6 V, with a quiescent current of 6.2 µA in non-sleep modes.
Sensors 2017, 17, 365
3 of 14
Sensors 2017, 17, 365 Sensors 2017, 17, 365
3 of 14 3 of 14
Figure 1. 1. Wearable sensor node block block diagram. Figure Wearable sensor Figure 1. Wearable sensor node node block diagram. diagram.
Figure 2. Wearable sensor node photograph: (Left) without XBee; (Right) with XBee. Figure 2. Wearable sensor node photograph: (Left) without XBee; (Right) with XBee. Figure 2. Wearable sensor node photograph: (Left) without XBee; (Right) with XBee.
The microcontroller manages the node operation, i.e., sensor acquisition; data collection and The microcontroller manages the node operation, i.e., sensor acquisition; data collection and storage; building of the data frame to benode sent by the radio frequency (RF) transceiver; sending of the Thebuilding microcontroller manages thebe i.e., sensor acquisition; data collection storage; of the data frame to sent operation, by the radio frequency (RF) transceiver; sending ofand the data frame to the of RFthe transceiver through an RS-232 protocol; and monitoring of the node state storage; building data frame to be sent by the radio frequency (RF) transceiver; sending data frame to the RF transceiver through an RS-232 protocol; and monitoring of the node state (energy mode, initialization, etc.). The node RF transceiver is an XBee module with a wire antenna to of the data frame to the RF etc.). transceiver through an RS-232 protocol; and monitoring the node (energy mode, initialization, The node RF transceiver is an XBee module with a wireofantenna to reduce the size. It includes the DM-24 firmware from DigiMesh that works in the 2.4 GHz state (energy mode, initialization, etc.). The node RF transceiver is an XBee module with wire reduce the size. It includes the DM-24 firmware from DigiMesh that works in the 2.4a GHz Industrial-Scientific-Medical (ISM) band with an IEEE 802.15.4 protocol running in the transceiver, antenna to reduce the size. It includes the with DM-24 from protocol DigiMeshrunning that works in the 2.4 GHz Industrial-Scientific-Medical (ISM) band anfirmware IEEE 802.15.4 in the transceiver, making compatible the connection of this with device toIEEE a deployed WSN and running allowingin drawing up of its Industrial-Scientific-Medical (ISM) band an 802.15.4 protocol the transceiver, making compatible the connection of this device to a deployed WSN and allowing drawing up of its communications structure. making compatible the connection of this device to a deployed WSN and allowing drawing up of its communications structure. The sensor node includes a novel and compact power supply system similar to those available communications structure. The sensor node includes a novel and compact power supply system similar to those available in smartphones and other portable consumer electronic devices. It consists of a MCP73833 battery The sensor node includes a novel and compact powerdevices. supply system similar those available in in smartphones and other portable consumer electronic It consists of atoMCP73833 battery charger circuit from Microchip, a rechargeable battery and two low dropout (LDO) regulators to smartphones and other portable consumer electronic devices. consists a MCP73833 charger charger circuit from Microchip, a rechargeable battery andIttwo low of dropout (LDO)battery regulators to provide the power voltage levels required by the different components. The selected battery is a 3.7 V circuit Microchip, a rechargeable battery and two lowcomponents. dropout (LDO) to provide providefrom the power voltage levels required by the different The regulators selected battery is a 3.7the V and 800 mAh Lithium Polymer (LiPo) LP-573442-1S-3 that meets the W-WSN node energy power voltage required by the (LiPo) differentLP-573442-1S-3 components. The selected is a 3.7 V node and 800 mAh and 800 mAhlevels Lithium Polymer that meetsbattery the W-WSN energy requirements with a small size. The voltage regulators provide stable, well-defined voltage levels Lithium Polymer (LiPo) LP-573442-1S-3 that meets the W-WSN node energy requirements with a levels small requirements with a small size. The voltage regulators provide stable, well-defined voltage from The the voltage decreasing batteryprovide voltage:stable, an MCP1725 LDO voltage voltage levels regulator (Microchip Technology size. regulators well-defined from the decreasing battery from the decreasing battery voltage: an MCP1725 LDO voltage regulator (Microchip Technology Inc., Chandler, AZ, USA) provides a 3 V constant voltage to power the XBee RF transceiver. The voltage: an MCP1725 LDO voltage regulator (Microchip Technology Inc.,the Chandler, AZ, USA) provides Inc., Chandler, AZ, USA) provides a 3 V constant voltage to power XBee RF transceiver. The second LDO regulator (TC1015 also from Microchip) provides a constant 2.5 V level to power the asecond 3 V constant voltage to(TC1015 power the XBee RFMicrochip) transceiver.provides The second LDO regulator (TC1015 also from LDO regulator also from a constant 2.5 V level to power the rest of node electronics. Microchip) a constant 2.5 V level to power the rest of node electronics. rest of nodeprovides electronics. Sensors are housed on a specific printed printed circuit board board (Figure 3) 3) connected to to the sensor sensor node Sensors are housed Sensors are housed on on aa specific specific printed circuit circuit board (Figure (Figure 3) connected connected to the the sensor node node through a pin in-line connector. This allows an easy replacement for damaged or obsolete sensors through through aa pin pin in-line in-line connector. connector. This This allows allows an an easy easy replacement replacement for for damaged damaged or or obsolete obsolete sensors sensors without modifying the the main node node architecture. architecture. The version herein presented measures of humidity, without The version version herein herein presented presented measures measures of of humidity, humidity, without modifying modifying the main main node architecture. The temperature and CO 2 concentration. A SHT11 (Sensirion AG, Zurich, Switzerland) provides the temperature A SHT11 SHT11 (Sensirion (Sensirion AG, AG, Zurich, Zurich, Switzerland) temperature and and CO CO22 concentration. concentration. A Switzerland) provides provides the the relative humidity humidityand andtemperature temperatureinformation, information,with withan anoperation operationrange rangeofof0% 0%toto100% 100%and and−−40 °C to relative 40 ◦°C C to relative humidity and temperature information, with an operation range of 0% to 100% and −40 to 124 °C, respectively. The gas detection is performed by a non-dispersive infrared IRC-A1 CO 2 sensor ◦ C, respectively. The gas detection is performed by a non-dispersive infrared IRC-A1 CO sensor 124 124 °C, respectively. The gas detection is performed by a non-dispersive infrared IRC-A1 CO22 sensor from Alphasense (Essex, UK) [12,16] in the range of 0 to 50,000 ppm to match safety applications. from Alphasense (Essex, UK) [12,16] in the range of 0 to 50,000 ppm to match safety applications. This sensor requires 2 Hz square wave excitation, a preheating time of 30 min and up to 40 s of This sensor requires 2 Hz square wave excitation, a preheating time of 30 min and up to 40 s of
Sensors 2017, 17, 365
4 of 14
Sensors 2017, 17, 365
4 of 14
from Alphasense (Essex, UK) [12,16] in the range of 0 to 50,000 ppm to match safety applications. response time. Our conditioning circuitry has been adapted from [17], fitand the up µCtoanalog-to-digital This sensor requires 2 Hz square wave excitation, a preheating time of 30 to min 40 s of response converter (ADC) input voltage requirements. time. Our conditioning circuitry has been adapted from [17], to fit the µC analog-to-digital converter The 2-Hz squarerequirements. signal required by the sensor is directly provided by an independent astable (ADC) input voltage circuit based operational (Maxim Integrated, San José, USA) instead of a The 2-Hzonsquare signal amplifier required MAX4038 by the sensor is directly provided by anCA, independent astable digital output from the microcontroller. Thus, the microcontroller can be set to a low-power mode, circuit based on operational amplifier MAX4038 (Maxim Integrated, San José, CA, USA) instead of the energy while Thus, the sensor keeps always can active to minimize its response areducing digital output from consumption, the microcontroller. the microcontroller be set to a low-power mode, time. reducing the energy consumption, while the sensor keeps always active to minimize its response time.
(a)
(b) Figure 3. Sensor connection: (a) photograph. Figure 3. Sensor board board and and node node connection: (a) block block diagram; diagram; (b) (b) photograph.
The three signals (reference, active and temperature voltages) in the IRC-A1 sensor must be three signals (reference, active and temperature voltages) in the IRC-A1 sensor must be used usedThe to calculate the CO2 concentration, by applying the equations given in the sensor datasheet. to calculate the CO2 concentration, by applying the equations given in the sensor datasheet. However, However, due to the complexity of the required equations and the use of a low-cost microcontroller due the complexity of thethe required equations and the use of a low-cost microcontroller as a sensor as a to sensor node manager, microcontroller includes look-up table (LUT) data that allows for a node manager, the microcontroller includes look-up table (LUT) data that allows for a coarse estimate coarse estimate of the measured CO2 levels, while an acoustic alarm using a buzzer included in the of the measured CO while an acoustic using a buzzer included in the board is activated 2 levels, board is activated when concentration exceedsalarm a predetermined level. when concentration exceeds a predetermined level. 3.2. Wearable Sensor Node Software 3.2. Wearable Sensor Node Software Figure 4 shows the operation flowchart of the wearable sensor node software. Once the power Figure 4 shows the operation flowchart of the wearable sensor node software. Once the power is is switched on for the very first time, the microcontroller is firstly configured using locally-stored switched on for the very first time, the microcontroller is firstly configured using locally-stored default default parameters: duty/sleep timing, operation frequencies, etc. Next, the RF transceiver is parameters: duty/sleep timing, operation frequencies, etc. Next, the RF transceiver is initialized initialized following a microcontroller request, initializing the universal asynchronous following a microcontroller request, initializing the universal asynchronous receiver-transmitter receiver-transmitter (UART), configuring the interrupt service routine (ISR) and starting the wireless (UART), configuring the interrupt service routine (ISR) and starting the wireless network joint process. network joint process. Network joining requires an XBee software reset, then a commissioning Network joining requires an XBee software reset, then a commissioning command is sent to the network command is sent to the network coordinator node to indicate that a new device is joining the network, and finally the transceiver is sent to sleep mode. The delays shown in the flowchart are required by the transceiver for a suitable processing of the commands.
Sensors 2017, 17, 365
5 of 14
coordinator node to indicate that a new device is joining the network, and finally the transceiver is sent to sleep mode. The delays shown in the flowchart are required by the transceiver for a suitable processing of2017, the17,commands. Sensors 365 5 of 14 Once both microcontroller and transceiver have been initialized, sensor input/outputs and Once both microcontroller and transceiver have been initialized, sensor input/outputs and peripherals (such as the analog-to-digital converter) are configured. Configuration is completed by peripherals (such as the analog-to-digital converter) are configured. Configuration is completed by enabling the microcontroller to receive interrupts from the transceiver, then driving the node to sleep enabling the microcontroller to receive interrupts from the transceiver, then driving the node to mode. Once node operation of aconsists main loop, the microcontroller is sleep fully mode.configured, Once fully the configured, the nodeconsists operation of a where main loop, where the awakened by an interrupt request the XBee. Next, increases microcontroller is awakened by from an interrupt request fromthe the operation XBee. Next, frequency the operation frequencyin order increases in order to time, reducethe the sensor acquisition sensor data are collected, to the in the to reduce the acquisition datatime, are the collected, compared to thecompared values stored values the LUT coordinator and sent to thethrough networkthe coordinator through the transceiver. Finally, the LUT and sentstored to theinnetwork transceiver. Finally, the operation frequency is operation frequency is decreased and the system goes back to sleep mode until a new XBee interrupt decreased and the system goes back to sleep mode until a new XBee interrupt is produced. is produced.
Figure 4. Flowchart of sensor node software operation. ISR: Interrupt service routine
Figure 4. Flowchart of sensor node software operation. ISR: Interrupt service routine. 4. Network Test
4. NetworkThe Testproposed system has been tested in the indoor deployment shown in Figure 5. Each room includes a static node (green), which sends parameters throughshown multi-hop to room The proposed system has been tested inambient the indoor deployment in transmission Figure 5. Each the node coordinator (orange, room #3), and then information is sent to a host device. includes a static node (green), which sends ambient parameters through multi-hop transmission to the node coordinator (orange, room #3), and then information is sent to a host device.
Sensors 2017, 17, 365
6 of 14 6 of 14 6 of 14
Room #1 #1 Room
Room #4 #4 Room
Sensors 2017, 17, 365 Sensors 2017, 17, 365
Figure 5. Indoor wearable wireless sensor network (W-WSN) deployment in the application test Figure 5. Indoor wearable wireless sensor network (W-WSN) deployment in the application test Figure 5. (dimensions: Indoor wearable scenario 50 m wireless × 15 m). sensor network (W-WSN) deployment in the application test scenario (dimensions: 50 m × 15 m). scenario (dimensions: 50 m × 15 m).
Wearable nodes (red, Figure 5) connect to the deployed network through the nearest fixed Wearable nodes (red, 5)5)connect to the deployed network throughthe thenearest nearestfixed fixed Wearable nodes (red,Figure Figure connect tothe theowner deployed network through device, therefore providing information about position and displacement. Because data device, therefore providing information position and displacement. Because data device, therefore providing information aboutthe the owner owner position and displacement. Because data transmissions are performed over shortabout distances, RF transmission power in mobile devices is set to transmissions are performed over short distances, RF power in mobile devices is setistoset transmissions are performed over short distances, RFtransmission transmission power in mobile devices the minimum available value, i.e., 10 dBm for XBee modules [18]. Node locations are upgraded the minimum value, i.e., for XBee modules [18]. Node locations are upgraded to the minimum available value,send i.e., 10 10 dBm dBm forhost XBee modules [18]. Node locations areevery upgraded every time theavailable network nodes data to the system. Busy periods are configured 15 s, every time the network nodes send data to the host system. Busy periods are configured every 1515 s, s, every timeallthe nodes send data thefrom hostthe system. Busy periods are configured every so the thenetwork monitored parameters areto sent wearable and static nodes via the host system so the all the monitored parameters are sent from the wearable and static nodes via the host system to all thethe processing application which processes thewearable data, activating suitable protocols insystem case ofto so the monitored parameters are sent from the and static nodes via the host the processing application processes the activating data, activating suitable protocols of unusual magnitude levels. theto processing application whichwhich processes the data, suitable protocols in caseinofcase unusual unusual magnitude levels. To check the correct operation of the proposed W-WSN, wearable nodes were positioned on magnitude levels. To check the correct operation of the proposed W-WSN, wearable nodes were positioned on volunteers along the monitored Figure W-WSN, 6 shows the measurements provided by one of To check moving the correct operation of thearea. proposed wearable nodes were positioned on volunteers moving alonga the monitored area. Figure days, 6 shows the measurements provided bypresents one of the volunteers wearing node for three consecutive for periods of 3.5–4.5 h. Figure 6a volunteers moving along the monitored area. Figure 6 shows the measurements provided by one the volunteers wearing a node for threewhile consecutive for periods of 3.5–4.5 h. Figure 6a presents measurement for Figure days, 6b shows the periods CO 2 concentration. 2 peaks of the volunteers results wearing a humidity, node for three consecutive days, for of 3.5–4.5 h.COFigure 6a measurement results for humidity, while Figure 6b shows the CO 2 concentration. CO2 peaks correspond to measurements when the user entered one of the dependencies occupied 26 presents measurement results for humidity, while Figure 6b shows the CO2 concentration. CO2bypeaks correspond students. to measurements when the user entered one of the dependencies occupied by 26 correspond to measurements when the user entered one of the dependencies occupied by 26 students. students. To validate the network link quality, the different packet error ratios (PER) have been To To validate the network link quality, the different packet error ratios (PER) have beenhave determined. validate the network link quality, the static different (PER) been determined. The results (Table 1) show a PER for nodespacket higher error than inratios the case of mobile nodes Thedetermined. results (Table 1) show a PER for static nodes higher than in the case of mobile nodes connected to results (Table 1) show PER static than the caseofofsight, mobile nodes connected toThe static devices, mainly duea to thefor fact thatnodes static higher links are in in non-line while the static devices,tomainly due to the fact due thatto static links arestatic in non-line of sight, while mobile nodes connected static mainly the fact that links are non-line of the sight, while the mobile nodes are devices, connected to the network with line of sight. Theinincrease of errors in packet aremobile connected to the network with line of sight. The increase of errors in packet transmission due nodes due are to connected to the network withtoline of sight. Thelocation. increase of errors in packet to transmission node reconnections is related changes in user node reconnections is related to changes in location. transmission due to node reconnections is user related to changes in user location. 46.5
R elative humhum idityidity (% )(% ) R elative
46.5 46 46 45.5
45.5 45 45 44.5
44.5 44 44 43.5
43.5 43 43 42.5 0 30 60 90 120 150 180 210 30 60 90 120 150 180 210 240 270 30 60 90 120 150 180 210 240 270 42.5 (min)210 240 270 30 60 90 120 150 180 210 240 270 0 30 60 90 120 150 180 210 30 60 90 120 Time 150 180
Time (a)(min)
(a)
Figure 6. Cont. Figure 6. Cont. Figure 6. Cont.
Sensors 2017, 17, 365 Sensors 2017, 17, 365
7 of 14 7 of 14
750
Sensors 2017, 17, 365 CO2 concentration (ppm ) (ppm ) CO2 concentration
700 750 650 700 600 650 550 600 500 550 450 500 400 0 450 400 0
7 of 14
30
60
90
120 150 180 210 30
60
90 120 150 180 210 240 270 30
30
60
90
120 150 180 210 30
60
90 120 150 180 210 240 270 30
Time (min)
(b)
60
90
120 150 180 210 240 270
60
90
120 150 180 210 240 270
(min) Figure 6. (a) Relative humidity acquired duringTime laboratory sessions; (b) CO2 concentration acquired Figure 6. (a) Relative humidity acquired during laboratory sessions; (b) CO2 concentration acquired (b) during these same periods. during these same periods. Figure 6. (a) Relative humidity acquired during laboratory sessions; (b) CO2 concentration acquired Table 1. Packet Error Ratio. during these same periods. Table 1. Packet Error Ratio.
Node
Table 1. Packet Error Ratio. Static Node Node
Rate (%) Rate 2.6 (%)
Mobile Node Connected 0.95 Node Rate (%) Static Nodeto Static Router 2.6 Mobile Node Reconnecting to Other Router 3.4 Static Node 2.6 Mobile Node Connected to Static Router 0.95 Mobile to Static Other Router Router 3.4 MobileNode NodeReconnecting Connected to 0.95 An important issue in theNode design of a wearable sensor node3.4 is its operating lifetime. To Mobile Reconnecting to wireless Other Router obtain a reliable node lifetime estimation, the evolution of the battery voltage and energy level in An important issue in the design of a wearable wireless sensor node is its operating lifetime. real operation conditions measured and numerically modeled. was verified that the An important issue inhas thebeen design of a wearable wireless sensor nodeFirst, is itsitoperating lifetime. To To obtain a reliable node lifetime estimation, the evolution of the battery voltage and energy level experimental power consumption is constant in each one of the node states along time. Then, to obtain a reliable node lifetime estimation, the evolution of the battery voltage and energy level in in real operation conditions has been measured and numerically modeled. First, it was verified that obtain the battery discharge for the 3.7 V, 800 mAh LiPo LP-573442-1S-3 battery, the current node real operation conditions has been measured and numerically modeled. First, it was verified that the theconsumption experimentalprofile powerwas consumption is constant in eachvoltage one of values the node states7). along time. Then, measured for several battery (Figure As the figure experimental power consumption is constant in each one of the node states along time. Then, to to obtain the battery discharge for the 3.7 V, 800 mAh LiPo LP-573442-1S-3 battery, the current node shows,the thebattery currentdischarge consumption is also independent of the battery outputbattery, voltage:the 5.5current mA in node sleep obtain for the 3.7 V, 800 mAh LiPo LP-573442-1S-3 consumption profile measured forfor several battery voltage values (Figure 7). As the shows, mode (mainly due was to was keeping the CO 2 sensor powered to voltage allow fast measurement without for consumption profile measured several battery values (Figure 7). As figure theneed figure theshows, current consumption is also independent of the battery output voltage: 5.5 mA in sleep mode warm-up time), 69 mA in acquisition mode and 60 mA in RF transmission. This is due to the use of the current consumption is also independent of the battery output voltage: 5.5 mA in sleep (mainly due to keeping the CO sensor powered to allow fast measurement without need for warm-up LDO voltage regulators to power the node at fixed voltages from the battery varying voltage. 2 mode (mainly due to keeping the CO2 sensor powered to allow fast measurement without need for time), 69 mAtime), in acquisition and 60 mA and in RF60transmission. This is dueThis to the voltage warm-up 69 mA in mode acquisition mode mA in RF transmission. is use dueof toLDO the use of regulators to power the node at fixed fromvoltages the battery voltage. LDO voltage regulators to power thevoltages node at fixed fromvarying the battery varying voltage.
Figure 7. Mobile node current consumption. Nodes are biased by a nominal 3.7 V battery, charged to 5 V (green), 4.2 V (red) and 3.4 V (blue). Figure Mobilenode nodecurrent currentconsumption. consumption. Nodes 3.73.7 VV battery, charged to to Figure 7. 7. Mobile Nodesare arebiased biasedby bya anominal nominal battery, charged 5 V (green), 4.2 V (red) and 3.4 V (blue). 5 V (green), 4.2 V (red) and 3.4 V (blue).
Sensors 2017, 17, 365
8 of 14
The battery discharge at constant current is shown in Figure 8: the battery voltage drastically Sensors 2017, 17, 365 8 of 14 drops after its value falls below 3.5 V. For the useful region, it is modeled according to The battery discharge at constant current is shown in Figure 8: the battery voltage drastically
13 10 v(tdrops 1.08 10−falls × t5 +3.5 1.493 ×the 10− ×region, t4 − 7.36 × 10−8 × t3 + 1.729 × 10−5 × t2 )= − after its × value below V. For useful it is modeled according to
−3 − 3.047 ×= 10 v in volts, × +(1.493 × 10 t×minutes − 7.36 )× 10 −1.08× × t10+ 4.125 Sensors 2017, 17, 365
10
×
− 3.047 × 10
× + 1.729 × × + 4.125 (v in volts, t in minutes)
(1)
8 (1) of 14
The battery discharge at constant current is shown in Figure 8: the battery voltage drastically drops after its value falls below 3.5 V. For the useful region, it is modeled according to × + 1.493 × 10 × − 7.36 × 10 × + 1.729 × = −1.08 × 10 10 × − 3.047 × 10 × + 4.125 (v in volts, t in minutes)
(1)
Figure 8. Experimental battery voltage (V) vs. time (min) for a constant 60 mA discharge.
Figure 8. Experimental battery voltage (V) vs. time (min) for a constant 60 mA discharge. Based on the results presented in Figure 8, the relationship between the remaining energy and the voltage provided by the battery can be obtained. The result is presented in Figure 9. This Based on the results presented in Figure 8, the relationship between the remaining energy behavior is approximately given by:
and the voltage provided by the battery can be obtained. The result is presented in Figure 9. This behavior is Figure 8. Experimental battery voltage (V) vs. time (min) for a constant 60 mA discharge. approximately given by: Based on the results presented in Figure 8, the relationship between the remaining energy and the voltage provided 5by the2 battery can be 6obtained. The result is6 presented in Figure 9. This C (v) = −1.243 × 10 × v + 1.026 × 10 × v − 2.073 × 10 (v in volts, C in mAmin) behavior is approximately given by:
(2)
Figure 9. Experimental (blue) and polynomial (red) fit for battery discharge (mAmin) vs. battery voltage. = −1.243 × 10 ×
+ 1.026 × 10 ×
− 2.073 × 10 (v in volts, C in mAmin)
(2)
Figure Experimental (blue) and polynomial (red) fit for discharge battery discharge battery On the 9.other hand, the equation of the node battery in each(mAmin) workingvs.cycle can be Figure 9. Experimental (blue) and polynomial (red) fit for battery discharge (mAmin) vs. voltage. obtained according to the process described in [15] as follows
battery voltage.
= −1.243 × 10 ×
C=
t1
t2
t3
i1 (t ) dt i2 ( t ) dt K i3 ( t )dt
t0 + 1.026 × 10 ×t 1 − 2.073 × 10t 2 (v in volts, C in mAmin)
(3) (2)
On the other hand, the equation of of the batterydischarge discharge in working each working can be On the other hand, the equation thenode node battery in each cycle cancycle be obtained according to the process described inin[15] follows obtained according to the process described [15] as follows C=
t0
t1
t
t
2 3 i (t ) dtZ t2 i2 ( t ) dt K Zi3 (tt3)dt t1 t2 i1 (t)dt + i2 (t)dt + K i3 (t)dt t1 t2
Z tC= 1
t0 1
(3)
(3)
where ix (t) represents the current in the three different node states: sleep, with time limits t0 and t1 ; measure, with time limits t1 and t2 ; and RF transmission, with time limits t2 and t3 . Since the current consumption in each state (see Figure 7) is constant, Equation (3) can be simplified to
Sensors 2017, 17, 365 Sensors 2017, 17, 365
9 of 14
9 of 14
where ix(t) represents the current in the three different node states: sleep, with time limits t0 and t1; measure, with time limits t1 and t2; and RF transmission, with time limits t2 and t3. Since the current − t2be (4) − t1 ) + i3(3) C Figure = i1 (t17)−ist0constant, (t3 can ) simplified to ) + i2 (t2 Equation consumption in each state (see C=i 1 t1the t 0battery i2 t 2discharge t1 i3 t 3asat 2function In this way, it is possible to estimate of time as
In this way, it is possible to estimate the battery discharge as a function of time as C (t) = −6.575 × t + 47998.38 (C in mAmin, t in minutes) = −6.575 × + 47998.38 (C in mAmin, t in minutes)
(4)
(5) (5)
Taking into account operatingvoltage voltageisislimited limited V, which corresponds Taking into accountthat thatthe the useful useful operating to to 3.53.5 V, which corresponds to a to a remaining energy of approximately 8000 mAmin (Figure 9), simulation results indicate a wearable remaining energy of approximately 8000 mAmin (Figure 9), simulation results indicate a wearable node operating 5.34 days daysexperimentally experimentallydetermined, determined, node operatinglife lifeofof5.07 5.07days days(Figure (Figure10), 10), close close to to the the 5.34 forfor a a duty cycle ofof 73.1 valuesensure ensureaasuitable suitableoperating operating duty cycle 73.1ms msand and1515s sofofworking workingcycle. cycle. These These values lifelife forfor a a wearable device, setting down downananappropriate appropriate schedule for battery recharge processes, wearable device,allowing allowing setting schedule for battery recharge processes, i.e., i.e.,this this operability time is long enough a portable system be charged on a basis. daily The basis. operability time is long enough for afor portable system to be to charged on a daily rechargeable system integrated in mobile sensor node node is essential for thisfor application and it isand an it The rechargeable system integrated in mobile sensor is essential this application improvement with respect to previous work. is an improvement with respect to previous work. Finally, Table 2 summarizes the main performances the proposed wearable node and Finally, Table 2 summarizes the main performances of theof proposed wearable node and compares compares them with other similar solutions. them with other similar solutions.
Figure Lifetimeestimation. estimation. Red Red line line indicates indicates the inin the battery. Figure 10.10.Lifetime theuseful usefulenergy energylimit limitstored stored the battery. Table 2. Sensor Nodes Comparison. Table 2. Sensor Nodes Comparison. Device Parameters
IRIS Crossbow IRIS Crossbow
Micaz Crossbow Micaz Crossbow
TelosB Crossbow TelosB Crossbow MSP430 16 bits MSP430 N/A 16 bits 1.8 mA N/A 5.1mA µA 1.8
Waspmote Libelium Waspmote Libelium ATMega1281 8 bits ATMega1281 8 MHz 8 bits 9 mA 8 MHz 62 µA 9 mA
This Work
Device Parameters This Work Processor Microcontroller ATMega1281 ATMega1281 PIC18F26J50 Processor N° Bits 8 bits 8 bits 8 bit Microcontroller ATMega1281 ATMega1281 PIC18F26J50 Frequency N/A N/A 8 MHz ◦ N Bits 8 bits 8 bits 8 bit Active Mode Current 8 mA 8 mA 7 mA Frequency N/A N/A 8 MHz Sleep ModeCurrent Current 8 µA 50 m 20–30 m 20–30 m 60–90 m 60–90 m Max. Tx Power 3 dBm 0 dBm 0 dBm 18 dBm 18 dBm Sensitivity −101 dBm −94 dBm −94 dBm −100 dBm −100 dBm Receive Mode 16 mA 19.7 mA 23 mA 57.08 mA 56.4 mA Max. Tx Power 3 dBm 0 dBm 0 dBm 18 dBm 18 dBm Transmission Current mA 17.4 mA 188 mA 69 mA Receive Mode 1617 mA 19.7 mA 23 N/A mA 57.08 mA 56.4 mA Sleep Mode NA 1 µA 1 µA 120 µA
