By Robert Dennis and Jon Edwards
E
photo courtesy of jon edwards
Considering Endoscopic Design A Snakebot Prototype
Digital Object Identifier 10.1109/MPUL.2013.2262141 Date of publication: 26 July 2013
30 ieee pulse ▼ JUlY/august 2013
ndoscope design is at the intersection of many disciplines, including robot design, computer science, material science, and medical devices. When considering design features, it is important to ensure that the device safely navigates the patient’s body and, once in position, performs the task required by the surgeon. Whether surgical access is achieved through existing orifices or by small incisions, the goal is to provide the surgeon with a stable platform within the patient from which to cut, suture, and grasp, while transmitting a clear image. The device described in this article can generate force at the tip of the snakebot in any direction without pressing against interstitial tissue to redirect the applied force. The prototype design is for the portion of the tool between the extracorporeal positioning structures and the end-effector and can be adapted to any hierarchal positioning mechanism. Other issues, such as the extracorporeal routing of the structures supporting this effector around the patient’s body and whether the robot is to be the surgeon’s slave or collaborator, although important considerations, are not addressed at this time.
History and the State of the Art The history of endoscopes extends many millennia, with Egyptian papyrus documenting endoscope use in 2,460 b.c. [1]. Modern endoscope history began in 1806 with Philip Bozzini’s lichleiter or light conductor [2]. Endoscopes were used by Hienz Kalk in the 1930s to diagnose liver and gallbladder disease and by Raoul Palmer in 1944 to perform gynecologic surgery. Commercial flexible gastroscopes became available in 1911 and semiflexible gastroscopes in 1930, but real progress in endoscope use and design began when Karl Storz created a cold light, an extracorporeal light source that routed illumination into the body without producing potentially damaging heat. In combination with the medical optics developed by Harold Hopkins, the endoscope became a viable tool. A camera for the gastroscope tip was developed in 1950 by Mutsuo Sugiura and was used to diagnose stomach ulcers. Hopkins introduced fiber optics in the 1950s, but they were unreliable and of low resolution until Hopkins and Storz optimized the optics, an evolution that enabled modern keyhole surgeries and earned Hopkins the Rumford Medal in 1984. Endoscope use changed with the introduction of robots, which never tire, do not tremble, and can report applied forces accurately.
2154-2287/13/$31.00©2013IEEE
The level of automation integration varies between platforms, from a generic holder attached to a standard human-operated endoscope to highly integrated master–slave systems in which the surgeon views a remote display and may not even be in the same city as the patient [3]. The design of the hyperredundant robot and its close relative, the continuum robot, also has a long history. The tensor arm manipulator implemented by Anderson [4] in 1967 demonstrated the utility of hyperredundant puppet robot design. Continuum robots share the tendon actuation with this design, but they rely on a flexible core for force redirection and, as they have fewer cables, are not as inherently compliant. Much of the current research is focused on developing capsule endoscopes, surgeon-interface refinements, and types of extracorporeal robot-positioning systems. The design of the portion of the endoscope in situ varies from rigid to flexible and has varying degrees of automation, depending on the system and application. Designs similar to this prototype are also being developed [5], but none of them combine the simplicity of design with the control offered by individually controlled axes.
Snakebot Design
The goal is to provide the surgeon with a stable platform within the patient from which to cut, suture, and grasp, while transmitting a clear image.
Endoscopes allow surgeons to navigate parts of the circulatory system and much of the body and to perform many surgeries more quickly, safely, and with better outcomes than traditional methods. Even so, endoscopes that are too flexible cannot provide sufficient stability to perform cutting and grasping, while those that are too stiff cannot navigate safely [6]. Some flexible endoscopes navigate by curling the tip of the scope, twisting the base to orient toward the goal, and then pushing the base forward. This procedure works as long as the forward force at the base is redirected along the shaft of the endoscope by the body of the patient toward the desired target, but it can damage the patient if the route requires a turn that is too sharp or impinges delicate tissue. Other designs incorporate multiple controllable segments (continuum manipulator endoscopes), but these still rely on contact with interstitial tissue to redirect force to the tip [7]. Our prototype design for the snakebot is a cannula that generates force and motion at the tip without requiring opposing force from interstitial tissue. The inspiration for the design came from the study of the human hand [8], which, like the snakebot, utilizes remote actuators. Force at the snakebot tip is transferred to the base through robot-segment compression and cable tension. The snakebot can push off against interstitial tissue, as is done with conventional endoscope navigation, or it can conform to lumen, creating minimal tissue stress. With the former paradigm, the tip flex is reduced and tip-force application is more deterministic. Because this endoscope is a cannula, that is, a delivery system for surgeons’ tools, there is a hollow core through which scalpels, grippers, or cameras can be exchanged in situ. The robot segments’ wall thickness can facilitate routing in the walls of fiber optics for illumination and endoscopic tip vision [9], delivery of saline, and vacuum lines, all without compromising surgical implement payload. This design is not new or unique, as it is simply a series of universal joints, but it improves upon previous implementations by maintaining a lumen though which tools can be interchanged and by using cables to replace hinges and pivots [10], [11]. The snakebot can self-propel through the patient’s body, producing motive force not only from the manipulation of the endoscope base but also through the serpentine motion of the snakebot’s body. Clearly, it can be dangerous to have a tool forcing itself through a patient and perhaps perforating an organ, so the robot is equipped with force sensors and proceeds under the surgeon’s direct control. Path planning becomes nontrivial as both modes of propulsion (serpentine and base generated) must consider potential interstitial damage. Snakelike robot designs vary widely, ranging from climbing robots to serpentine rolling robots to elephant trunks [12]. Actuator type and placement vary as well, from external pneumatics to internal (to the snakebot) motor or memory metal. This particular design is a puppet controlled by cables (tendons) attached to robot segments at JUlY/august 2013 ▼ ieee pulse 31
controlled by four Arduino 328p single-board computers (SBCs) and four Arduino motor shields. One Arduino 2860 SBC connects four 328p via the two-wire interface (also known as I2C) protocol. The 2860 also communicates with a PC using a universal serial bus. Each 328p has a joystick used in training mode to move the two axes it controls, and the 2560 can store the axes’ configuration Prototype Considerations sequences in electrically eraseThe snakebot prototype is a able programmable read-only puppet consisting of stacked memory (EEPROM) for subsegments and cables (also sequent repetition. There are known as tendons). Each segno axis encoders; the stepper ment comprises a pair of ormotor count provides posithogonal hemicylindrical cams tion data. and is held in place by structural The dimensions of the cam cables and positioned by condetermine how sharp a bend trol cables. The control cables the snakebot can navigate. are routed through the snakeIn the prototype, the radius bot body to the base across a Figure 1 The snakebot prototype. (Photo courtesy of Jon of the hemicylindrical cam force sensor and to a stepper Edwards.) is equal to the radius of the motor (Figure 1). The robot was snakebot segment; a larger cam radius would increase the minidesigned using Solidworks CAD software, and the segments were mum radius bend navigable but, for a given applied tip force, it formed by force deposition modeling. The snakebot base is bolted would decrease the tension on cables and on subsequent segto an aluminum plate on which the eight stepper motors conments through which the cables pass. The length of the segment trolling the eight segments are mounted. The eight steppers are one end and to a stepper motor at the other. The advantages are that the motors’ ferromagnetic parts are as remote as necessary to allow the effector to be used in magnetic resonance imaging (MRI) or X-ray machinery, and the effector can be made as small as necessary to perform the surgery and still generate significant force at the tip.
Figure 2 The baseplate on which the snakebot is built. (Photo courtesy of Jon Edwards.) 32 ieee pulse ▼ JUlY/august 2013
Figure 3 The force sensor block. (Photo courtesy of Jon Edwards.)
also proportionately affects the minimum bend radius, with increased segment length increasing the lever arm of each segment joint, thereby increasing the cable tension per tip force. The snakebot’s dimensions can be customized for a given application, with a tradeoff of flexibility and complexity. Segments do not have to be identical in dimension, so the tip can be composed of shorter, more dexterous segments, while the body can be composed of longer, more stable segments. Routing channels for the control cables must ensure that the total cable length (measured from segment to stepper and back to the other side of the segment) remains constant across the range of axis movement. Otherwise, the movement of one axis would create hysteresis slop in the control cable of the segment being moved and create tension (or slack) in all control cables of segments between the axis being moved and the snakebot’s base. In the prototype, a force sensor that monitors the cable tension actuating each axis (Figures 2 and 3) is built into the baseplate supporting the effector. There are several reasons for collecting these data. First, it is a way to generate axis limit data, since the limit count of the stepper motor associated with a given axis changes as axes between that axis and the base move. Second, it provides improved effector control and increased safety [13]. Third, it allows for detection of interstitial tissue type by palpitation; bone will offer little deflection/force, while soft tissues will offer more. Finally, it allows for active position maintenance by having the steppers increase cable tension in response to applied force. Serpentine motion is rudimentary, with only eight segments (four segments per degree of freedom). The prototype has a 2-in diameter with a 1-in hollow core; it is 9.6 in long and can curve into a 5-in circle. The payload is 25% of the snake’s axial cross section, which will be increased in subsequent versions by thinning segment walls. Heuristically, a snakebot with thinner segment walls is less able to withstand lateral stress, flexes more to longitudinal stress, and is generally less stable than thicker-walled versions. This aspect is complicated by the probable effect of implementing the snakebot in nonferrous metal [titanium–copper (Ti–Cu) alloys are suitable], which will have different stress characteristics and therefore result in different snakebot performance. Tools can be exchanged in situ through the hollow core, and hollow channels in segment walls can facilitate nonreplaceable flexible tools (lights, suction, fiber optics, and saline). When unsupported by interstitia, the snakebot tip will flex if subjected to force, but it can also actively compensate for applied forces. Furthermore, this design can generate force in any direction, so it could wrap around an organ and operate on its distal surface or wrap around tissue and pull like an octopus. The movement of a segment changes the movement limits of all the subsequent segments, so the coordinated motion of multiple axes is necessary to maintain orientation of the tip. The kinematics is complicated by the flex of the snakebot due to several factors, including cable stretch, stepper flex, and segment deformation, which change with the snakebot’s position [14]. Deterministic control may be inherently impossible, but heuristics, fuzzy logic, or expert system control may compensate for physical limitations [15], [16].
Figure 4 The snakebot and baseplate. (Photo courtesy of Jon Edwards.)
Figure 5 The snakebot mounted on a sled. (Photo courtesy of Jon Edwards.)
Figure 6 The snakebot beginning a maze. (Photo courtesy of Jon Edwards.) JUlY/august 2013 ▼ ieee pulse 33
The snakebot can self-propel through the patient’s body, producing motive force not only from the manipulation of the endoscope base but also through the serpentine motion of the snakebot’s body.
Figure 7 The snakebot completing a maze. (Photo courtesy of Jon Edwards.)
Proof To prove that the snakebot is able to contort and then generate a tip force, the snakebot baseplate (Figure 4) was mounted on a sled (Figure 5) so that it could translate in the z axis. The snakebot is then threaded through a three-dimensional maze (Figures 6 and 7). The movements of the snakebot are documented in two videos (see https://www.youtube.com/ watch?v=mSu23Sg4Imw and https://www.youtube.com/ watch?v=UyPlOwdfA5E).
Future Considerations The research proposals for this device describe how a pair of snakebots can operate together, resembling a caduceus. This paradigm offers increased stability and strength, the possibility of stereo vision, two-handed tool usage, and the ability for one snakebot to use the second to facilitate serpentine propulsion and force generation, minimizing unintended stress against interstitial tissue. The caduceus paradigm also enables active compensation, whereby the flex generated by the active snakebot (as when cutting or pulling) is compensated for by the motor tension in the stationary snakebot. Additional stability can be achieved by 34 ieee pulse ▼ JUlY/august 2013
adding an inflatable ring along the length of the snakebot so that only the portion of the endoscope distal to the ring is subject to force flex. Each axis’ control cable crosses a force sensor, which enables it to palpitate to infer tissue type (by deflection/force) and provides a potential safeguard against tissue damage by inadvertent generation of excessive force. A force/torque sensor should be inserted between the plate holding the stepper motors and the snakebot base to collect axial snakebot force/torque data [13]. The interface presented to the surgeon is to be a drive-fromthe-tip perspective, with the body of the snakebot performing a follow-the-leader paradigm. Path planning will be complicated by the performance of serpentine motion for self-propagation, plotting safe hard points against which the endoscope can press when redirecting force, and the necessity of monitoring force sensors to prevent collateral damage. The movement of the tip must conform to the surgeon’s controls, but the movement of the body will be a function of positioning constraints, interstitial tissue limitations, and the need to generate force in useful vectors. Implementing the design in a nonferromagnetic alloy to allow for use in MRI machinery at a size useful to surgeons remains a challenge. Sintered Ti–Cu alloys present a cost-effective manufacturing possibility. Routing of fiber-optic cables within the segment walls will allow both light and vision at the robot’s tip. Since the tip is a ring, a form of raster scanning can be implemented. By arranging multiple fibers oriented axially at the tip of the robot and postprocessing successive snapshots, an image of higher density can be generated [9]. The impact of this design on the surgeon’s ability to perform complex surgeries with minimal impact on the patient is potentially large, as it provides the ability to generate tip force in any direction without requiring contact with interstitial tissue to redirect force. The current design is too nondeterministic for conventional kinematics, as it has excessive lateral flex from segment compression and cable stretch. However, when the shortcomings are compensated for, either by use in a caduceus paradigm or by heuristics, this platform enables access of
Endoscope use changed with the introduction of robots, which never tire, do not tremble, and can report applied forces accurately.
distal surgical site surfaces with proximal access ports, stereo vision and two-handed tool use (again under the caduceus paradigm), and in situ tool exchange. The incorporated force sensors, although not part of the snakebot per se, allow for tissue identification through palpitation and for future haptic control. In short, this design is a good start, a vector toward a more robust and useful surgical tool set. Robert Dennis ([email protected]) and Jon Edwards ([email protected]) are with the University of North Carolina, Chapel Hill.
References [1] C. Nezhat, Nezhat’s History of Endoscopy: A Historical Analysis of Endoscopy’s Ascension Since Antiquity. Endo Press, 2011. [2] Society of Laparoendoscopic Surgeons. Nezhat’s History of Endoscopy [Online]. Available: http://laparoscopy.blogs.com/ endoscopyhistory/chapter_06/ [3] Da Vinci Robotics. (2013). Da Vinci Surgery [Online]. Available: http://www.davincisurgery.com/ [4] V. C. Anderson and R. C. Horn, “Tensor arm manipulator design,” Trans. ASME, vol. 67-DE-57, no. 8, pp. 1–12, 1967. [5] i-Snake Surgical Robot for Minimally Invasive Surgery [Online]. Available: http://www1.imperial.ac.uk/medicine/research/ researchthemes/healthtechnologies/surgicaltechnologies/ isnake/. [6] J. Burdick and W. Grundfest, “The development of a robotic endoscope” in Proc. 1995 IEEE/RSJ Int. Conf. Intelligent Human Robot Interaction and Cooperative Robots, Robots and Systems 95, vol. 2, pp. 162–171. [7] R. J. Webster III and B. A. Jones, “Design and kinematic modeling of constant curvature continuum robots: A review,” Int. J. Robot. Res., vol. 29, no. 13, pp. 1661–1683, 2010. [8] J. A. Rieffel, F. J. Valero-Cuevas, and H. Lipson, “Morphological communication: Exploiting coupled dynamics in a complex mechanical structure to achieve locomotion,” J. R. Soc. Interface, vol. 45, pp. 613–621, Apr. 2010.
[9] D. R. Rivera, “Compact and flexible raster scanning multiphoton endoscope capable of imaging unstained tissue,” Proc. Natl. Acad. Sci. USA, vol. 108, no. 43, pp. 17598–17603, Oct. 2011. [10] M. Wada, “Flexibly foldable arm,” U.S. patent 4 685 349, Aug. 11, 1987. [11] A. Belson, “Steerable endoscope and improved method of insertion,” U.S. patent 8 062 212, Nov. 22, 2011. [12] M. W. Hannan and I. D. Walker, “The ‘elephant trunk’ manipulator, design and implementation,” in Proc. 2001 IEEE/ASME Int. Conf. Advanced Intelligent Mechatronics, vol. 1, pp. 14–19, doi: 10.1109/AIM.2001.936423. [13] K. W. Eichhorn, K. Tingelhoff, I. Wagner, R. Westphal, M. Rilk, M. E. Kunkel, F. M. Wahl, and F. Bootz, “Evaluation of force data with a force/torque sensor during FESS: A step towards robotassisted surgery,” (in German), HNO, vol. 56, no. 8, pp. 789–794, Aug. 2008. [14] R. H. Struges, Jr., “A flexible, tendon-controlled device for endoscopy,” Int. J. Robot. Res., vol. 12, no. 2, pp. 121–31, 1993. [15] W. T. Clement and R. M. Iñigo, “Design of a snake-like manipulator,” J. Robot. Auton. Syst., vol. 6, no. 3, pp. 265–282, July 1990. [16] C.-C. Lin, J.-H. Chuang, and C.-T. Hsieh, “A path planning algorithm using generalized potential model for hyper-redundant robots with 2-DOF joints,” Int. J. Adv. Robot. Syst., vol. 8, no. 2, pp. 49–58, June 2011.
JUlY/august 2013 ▼ ieee pulse 35
E
photo courtesy of jon edwards
Considering Endoscopic Design A Snakebot Prototype
Digital Object Identifier 10.1109/MPUL.2013.2262141 Date of publication: 26 July 2013
30 ieee pulse ▼ JUlY/august 2013
ndoscope design is at the intersection of many disciplines, including robot design, computer science, material science, and medical devices. When considering design features, it is important to ensure that the device safely navigates the patient’s body and, once in position, performs the task required by the surgeon. Whether surgical access is achieved through existing orifices or by small incisions, the goal is to provide the surgeon with a stable platform within the patient from which to cut, suture, and grasp, while transmitting a clear image. The device described in this article can generate force at the tip of the snakebot in any direction without pressing against interstitial tissue to redirect the applied force. The prototype design is for the portion of the tool between the extracorporeal positioning structures and the end-effector and can be adapted to any hierarchal positioning mechanism. Other issues, such as the extracorporeal routing of the structures supporting this effector around the patient’s body and whether the robot is to be the surgeon’s slave or collaborator, although important considerations, are not addressed at this time.
History and the State of the Art The history of endoscopes extends many millennia, with Egyptian papyrus documenting endoscope use in 2,460 b.c. [1]. Modern endoscope history began in 1806 with Philip Bozzini’s lichleiter or light conductor [2]. Endoscopes were used by Hienz Kalk in the 1930s to diagnose liver and gallbladder disease and by Raoul Palmer in 1944 to perform gynecologic surgery. Commercial flexible gastroscopes became available in 1911 and semiflexible gastroscopes in 1930, but real progress in endoscope use and design began when Karl Storz created a cold light, an extracorporeal light source that routed illumination into the body without producing potentially damaging heat. In combination with the medical optics developed by Harold Hopkins, the endoscope became a viable tool. A camera for the gastroscope tip was developed in 1950 by Mutsuo Sugiura and was used to diagnose stomach ulcers. Hopkins introduced fiber optics in the 1950s, but they were unreliable and of low resolution until Hopkins and Storz optimized the optics, an evolution that enabled modern keyhole surgeries and earned Hopkins the Rumford Medal in 1984. Endoscope use changed with the introduction of robots, which never tire, do not tremble, and can report applied forces accurately.
2154-2287/13/$31.00©2013IEEE
The level of automation integration varies between platforms, from a generic holder attached to a standard human-operated endoscope to highly integrated master–slave systems in which the surgeon views a remote display and may not even be in the same city as the patient [3]. The design of the hyperredundant robot and its close relative, the continuum robot, also has a long history. The tensor arm manipulator implemented by Anderson [4] in 1967 demonstrated the utility of hyperredundant puppet robot design. Continuum robots share the tendon actuation with this design, but they rely on a flexible core for force redirection and, as they have fewer cables, are not as inherently compliant. Much of the current research is focused on developing capsule endoscopes, surgeon-interface refinements, and types of extracorporeal robot-positioning systems. The design of the portion of the endoscope in situ varies from rigid to flexible and has varying degrees of automation, depending on the system and application. Designs similar to this prototype are also being developed [5], but none of them combine the simplicity of design with the control offered by individually controlled axes.
Snakebot Design
The goal is to provide the surgeon with a stable platform within the patient from which to cut, suture, and grasp, while transmitting a clear image.
Endoscopes allow surgeons to navigate parts of the circulatory system and much of the body and to perform many surgeries more quickly, safely, and with better outcomes than traditional methods. Even so, endoscopes that are too flexible cannot provide sufficient stability to perform cutting and grasping, while those that are too stiff cannot navigate safely [6]. Some flexible endoscopes navigate by curling the tip of the scope, twisting the base to orient toward the goal, and then pushing the base forward. This procedure works as long as the forward force at the base is redirected along the shaft of the endoscope by the body of the patient toward the desired target, but it can damage the patient if the route requires a turn that is too sharp or impinges delicate tissue. Other designs incorporate multiple controllable segments (continuum manipulator endoscopes), but these still rely on contact with interstitial tissue to redirect force to the tip [7]. Our prototype design for the snakebot is a cannula that generates force and motion at the tip without requiring opposing force from interstitial tissue. The inspiration for the design came from the study of the human hand [8], which, like the snakebot, utilizes remote actuators. Force at the snakebot tip is transferred to the base through robot-segment compression and cable tension. The snakebot can push off against interstitial tissue, as is done with conventional endoscope navigation, or it can conform to lumen, creating minimal tissue stress. With the former paradigm, the tip flex is reduced and tip-force application is more deterministic. Because this endoscope is a cannula, that is, a delivery system for surgeons’ tools, there is a hollow core through which scalpels, grippers, or cameras can be exchanged in situ. The robot segments’ wall thickness can facilitate routing in the walls of fiber optics for illumination and endoscopic tip vision [9], delivery of saline, and vacuum lines, all without compromising surgical implement payload. This design is not new or unique, as it is simply a series of universal joints, but it improves upon previous implementations by maintaining a lumen though which tools can be interchanged and by using cables to replace hinges and pivots [10], [11]. The snakebot can self-propel through the patient’s body, producing motive force not only from the manipulation of the endoscope base but also through the serpentine motion of the snakebot’s body. Clearly, it can be dangerous to have a tool forcing itself through a patient and perhaps perforating an organ, so the robot is equipped with force sensors and proceeds under the surgeon’s direct control. Path planning becomes nontrivial as both modes of propulsion (serpentine and base generated) must consider potential interstitial damage. Snakelike robot designs vary widely, ranging from climbing robots to serpentine rolling robots to elephant trunks [12]. Actuator type and placement vary as well, from external pneumatics to internal (to the snakebot) motor or memory metal. This particular design is a puppet controlled by cables (tendons) attached to robot segments at JUlY/august 2013 ▼ ieee pulse 31
controlled by four Arduino 328p single-board computers (SBCs) and four Arduino motor shields. One Arduino 2860 SBC connects four 328p via the two-wire interface (also known as I2C) protocol. The 2860 also communicates with a PC using a universal serial bus. Each 328p has a joystick used in training mode to move the two axes it controls, and the 2560 can store the axes’ configuration Prototype Considerations sequences in electrically eraseThe snakebot prototype is a able programmable read-only puppet consisting of stacked memory (EEPROM) for subsegments and cables (also sequent repetition. There are known as tendons). Each segno axis encoders; the stepper ment comprises a pair of ormotor count provides posithogonal hemicylindrical cams tion data. and is held in place by structural The dimensions of the cam cables and positioned by condetermine how sharp a bend trol cables. The control cables the snakebot can navigate. are routed through the snakeIn the prototype, the radius bot body to the base across a Figure 1 The snakebot prototype. (Photo courtesy of Jon of the hemicylindrical cam force sensor and to a stepper Edwards.) is equal to the radius of the motor (Figure 1). The robot was snakebot segment; a larger cam radius would increase the minidesigned using Solidworks CAD software, and the segments were mum radius bend navigable but, for a given applied tip force, it formed by force deposition modeling. The snakebot base is bolted would decrease the tension on cables and on subsequent segto an aluminum plate on which the eight stepper motors conments through which the cables pass. The length of the segment trolling the eight segments are mounted. The eight steppers are one end and to a stepper motor at the other. The advantages are that the motors’ ferromagnetic parts are as remote as necessary to allow the effector to be used in magnetic resonance imaging (MRI) or X-ray machinery, and the effector can be made as small as necessary to perform the surgery and still generate significant force at the tip.
Figure 2 The baseplate on which the snakebot is built. (Photo courtesy of Jon Edwards.) 32 ieee pulse ▼ JUlY/august 2013
Figure 3 The force sensor block. (Photo courtesy of Jon Edwards.)
also proportionately affects the minimum bend radius, with increased segment length increasing the lever arm of each segment joint, thereby increasing the cable tension per tip force. The snakebot’s dimensions can be customized for a given application, with a tradeoff of flexibility and complexity. Segments do not have to be identical in dimension, so the tip can be composed of shorter, more dexterous segments, while the body can be composed of longer, more stable segments. Routing channels for the control cables must ensure that the total cable length (measured from segment to stepper and back to the other side of the segment) remains constant across the range of axis movement. Otherwise, the movement of one axis would create hysteresis slop in the control cable of the segment being moved and create tension (or slack) in all control cables of segments between the axis being moved and the snakebot’s base. In the prototype, a force sensor that monitors the cable tension actuating each axis (Figures 2 and 3) is built into the baseplate supporting the effector. There are several reasons for collecting these data. First, it is a way to generate axis limit data, since the limit count of the stepper motor associated with a given axis changes as axes between that axis and the base move. Second, it provides improved effector control and increased safety [13]. Third, it allows for detection of interstitial tissue type by palpitation; bone will offer little deflection/force, while soft tissues will offer more. Finally, it allows for active position maintenance by having the steppers increase cable tension in response to applied force. Serpentine motion is rudimentary, with only eight segments (four segments per degree of freedom). The prototype has a 2-in diameter with a 1-in hollow core; it is 9.6 in long and can curve into a 5-in circle. The payload is 25% of the snake’s axial cross section, which will be increased in subsequent versions by thinning segment walls. Heuristically, a snakebot with thinner segment walls is less able to withstand lateral stress, flexes more to longitudinal stress, and is generally less stable than thicker-walled versions. This aspect is complicated by the probable effect of implementing the snakebot in nonferrous metal [titanium–copper (Ti–Cu) alloys are suitable], which will have different stress characteristics and therefore result in different snakebot performance. Tools can be exchanged in situ through the hollow core, and hollow channels in segment walls can facilitate nonreplaceable flexible tools (lights, suction, fiber optics, and saline). When unsupported by interstitia, the snakebot tip will flex if subjected to force, but it can also actively compensate for applied forces. Furthermore, this design can generate force in any direction, so it could wrap around an organ and operate on its distal surface or wrap around tissue and pull like an octopus. The movement of a segment changes the movement limits of all the subsequent segments, so the coordinated motion of multiple axes is necessary to maintain orientation of the tip. The kinematics is complicated by the flex of the snakebot due to several factors, including cable stretch, stepper flex, and segment deformation, which change with the snakebot’s position [14]. Deterministic control may be inherently impossible, but heuristics, fuzzy logic, or expert system control may compensate for physical limitations [15], [16].
Figure 4 The snakebot and baseplate. (Photo courtesy of Jon Edwards.)
Figure 5 The snakebot mounted on a sled. (Photo courtesy of Jon Edwards.)
Figure 6 The snakebot beginning a maze. (Photo courtesy of Jon Edwards.) JUlY/august 2013 ▼ ieee pulse 33
The snakebot can self-propel through the patient’s body, producing motive force not only from the manipulation of the endoscope base but also through the serpentine motion of the snakebot’s body.
Figure 7 The snakebot completing a maze. (Photo courtesy of Jon Edwards.)
Proof To prove that the snakebot is able to contort and then generate a tip force, the snakebot baseplate (Figure 4) was mounted on a sled (Figure 5) so that it could translate in the z axis. The snakebot is then threaded through a three-dimensional maze (Figures 6 and 7). The movements of the snakebot are documented in two videos (see https://www.youtube.com/ watch?v=mSu23Sg4Imw and https://www.youtube.com/ watch?v=UyPlOwdfA5E).
Future Considerations The research proposals for this device describe how a pair of snakebots can operate together, resembling a caduceus. This paradigm offers increased stability and strength, the possibility of stereo vision, two-handed tool usage, and the ability for one snakebot to use the second to facilitate serpentine propulsion and force generation, minimizing unintended stress against interstitial tissue. The caduceus paradigm also enables active compensation, whereby the flex generated by the active snakebot (as when cutting or pulling) is compensated for by the motor tension in the stationary snakebot. Additional stability can be achieved by 34 ieee pulse ▼ JUlY/august 2013
adding an inflatable ring along the length of the snakebot so that only the portion of the endoscope distal to the ring is subject to force flex. Each axis’ control cable crosses a force sensor, which enables it to palpitate to infer tissue type (by deflection/force) and provides a potential safeguard against tissue damage by inadvertent generation of excessive force. A force/torque sensor should be inserted between the plate holding the stepper motors and the snakebot base to collect axial snakebot force/torque data [13]. The interface presented to the surgeon is to be a drive-fromthe-tip perspective, with the body of the snakebot performing a follow-the-leader paradigm. Path planning will be complicated by the performance of serpentine motion for self-propagation, plotting safe hard points against which the endoscope can press when redirecting force, and the necessity of monitoring force sensors to prevent collateral damage. The movement of the tip must conform to the surgeon’s controls, but the movement of the body will be a function of positioning constraints, interstitial tissue limitations, and the need to generate force in useful vectors. Implementing the design in a nonferromagnetic alloy to allow for use in MRI machinery at a size useful to surgeons remains a challenge. Sintered Ti–Cu alloys present a cost-effective manufacturing possibility. Routing of fiber-optic cables within the segment walls will allow both light and vision at the robot’s tip. Since the tip is a ring, a form of raster scanning can be implemented. By arranging multiple fibers oriented axially at the tip of the robot and postprocessing successive snapshots, an image of higher density can be generated [9]. The impact of this design on the surgeon’s ability to perform complex surgeries with minimal impact on the patient is potentially large, as it provides the ability to generate tip force in any direction without requiring contact with interstitial tissue to redirect force. The current design is too nondeterministic for conventional kinematics, as it has excessive lateral flex from segment compression and cable stretch. However, when the shortcomings are compensated for, either by use in a caduceus paradigm or by heuristics, this platform enables access of
Endoscope use changed with the introduction of robots, which never tire, do not tremble, and can report applied forces accurately.
distal surgical site surfaces with proximal access ports, stereo vision and two-handed tool use (again under the caduceus paradigm), and in situ tool exchange. The incorporated force sensors, although not part of the snakebot per se, allow for tissue identification through palpitation and for future haptic control. In short, this design is a good start, a vector toward a more robust and useful surgical tool set. Robert Dennis ([email protected]) and Jon Edwards ([email protected]) are with the University of North Carolina, Chapel Hill.
References [1] C. Nezhat, Nezhat’s History of Endoscopy: A Historical Analysis of Endoscopy’s Ascension Since Antiquity. Endo Press, 2011. [2] Society of Laparoendoscopic Surgeons. Nezhat’s History of Endoscopy [Online]. Available: http://laparoscopy.blogs.com/ endoscopyhistory/chapter_06/ [3] Da Vinci Robotics. (2013). Da Vinci Surgery [Online]. Available: http://www.davincisurgery.com/ [4] V. C. Anderson and R. C. Horn, “Tensor arm manipulator design,” Trans. ASME, vol. 67-DE-57, no. 8, pp. 1–12, 1967. [5] i-Snake Surgical Robot for Minimally Invasive Surgery [Online]. Available: http://www1.imperial.ac.uk/medicine/research/ researchthemes/healthtechnologies/surgicaltechnologies/ isnake/. [6] J. Burdick and W. Grundfest, “The development of a robotic endoscope” in Proc. 1995 IEEE/RSJ Int. Conf. Intelligent Human Robot Interaction and Cooperative Robots, Robots and Systems 95, vol. 2, pp. 162–171. [7] R. J. Webster III and B. A. Jones, “Design and kinematic modeling of constant curvature continuum robots: A review,” Int. J. Robot. Res., vol. 29, no. 13, pp. 1661–1683, 2010. [8] J. A. Rieffel, F. J. Valero-Cuevas, and H. Lipson, “Morphological communication: Exploiting coupled dynamics in a complex mechanical structure to achieve locomotion,” J. R. Soc. Interface, vol. 45, pp. 613–621, Apr. 2010.
[9] D. R. Rivera, “Compact and flexible raster scanning multiphoton endoscope capable of imaging unstained tissue,” Proc. Natl. Acad. Sci. USA, vol. 108, no. 43, pp. 17598–17603, Oct. 2011. [10] M. Wada, “Flexibly foldable arm,” U.S. patent 4 685 349, Aug. 11, 1987. [11] A. Belson, “Steerable endoscope and improved method of insertion,” U.S. patent 8 062 212, Nov. 22, 2011. [12] M. W. Hannan and I. D. Walker, “The ‘elephant trunk’ manipulator, design and implementation,” in Proc. 2001 IEEE/ASME Int. Conf. Advanced Intelligent Mechatronics, vol. 1, pp. 14–19, doi: 10.1109/AIM.2001.936423. [13] K. W. Eichhorn, K. Tingelhoff, I. Wagner, R. Westphal, M. Rilk, M. E. Kunkel, F. M. Wahl, and F. Bootz, “Evaluation of force data with a force/torque sensor during FESS: A step towards robotassisted surgery,” (in German), HNO, vol. 56, no. 8, pp. 789–794, Aug. 2008. [14] R. H. Struges, Jr., “A flexible, tendon-controlled device for endoscopy,” Int. J. Robot. Res., vol. 12, no. 2, pp. 121–31, 1993. [15] W. T. Clement and R. M. Iñigo, “Design of a snake-like manipulator,” J. Robot. Auton. Syst., vol. 6, no. 3, pp. 265–282, July 1990. [16] C.-C. Lin, J.-H. Chuang, and C.-T. Hsieh, “A path planning algorithm using generalized potential model for hyper-redundant robots with 2-DOF joints,” Int. J. Adv. Robot. Syst., vol. 8, no. 2, pp. 49–58, June 2011.
JUlY/august 2013 ▼ ieee pulse 35
