1 Biomimetic Design James Tacey, Aditi Shinde Abstract—Robot locomotion is the name associated with robotic movement. This is the area of research that focuses on robots moving themselves from one location or space to another, as well as how they handle various terrains and their effects. For a robot to decide how, when and where to move autonomously is the foremost aim in mimetic locomotion. Synchronization of various parts and joints while moving around a place is one of the complex part of robot design. designers are now providing backup systems and tolerant systems in place on robots. These systems will detect when the robot is damaged or faulty, as well as which parts, and take steps to bring redundant systems online, compensate for the damage, or even repair itself. Index Terms—Biomimetic, locomotion, degrees of freedom, actuators, autonomous. II. ARTIFICIAL INTELLIGENCE:REVIEW STAGE I. INTRODUCTION The ability to adapt to new situation rapidly is an extremely beneficial tool. Much like a biological creature’s ability to learn, research into artificial intelligence has been attempting to give robots this same capability. Currently, as robots are only able to respond to pre-programmed stimuli, they are very limited in what they are capable of without a human controller. The ability to learn and adapt to new experiences would give robots the capability to respond to new situations and become more efficient at tasks. By using positive and negative feedback loops simple artificial intelligence can be achieved, but there is still a very large potential in this area. For years biological creatures have been adapting, traversing, and existing in a wide variety of natural environments without outside influence by humans. In comparison, robots in the past have been limited in capability by the limitations imposed by the technology available. More recently, robotic design has shifted towards mimicry of animals and their capabilities. By drawing insight from animal behavior and systems, more recent generations of robots have been able to be more efficient, adaptive, and to some extent, even capable of learning. Robotic Motion Mechanisms: Biomimetic Locomotion From Biology: As biomimetic locomotion is largely imitation and duplication of structures found in nature, it is important to fully understand what is being imitated and the biological mimicry behind it. Neurological control, redundant systems, and the ability to learn are all areas that robots can benefit from. Examined are a few examples. Neural Networks: Through the study of how animals and people control their various muscles, new systems can be developed that provide more robust control systems. For example, by developing neural networks in software, navigation systems can be made that are similar to a biological thought process. Because of this, robots can be developed that are more robust, and can respond to a wider variety of input stimuli. Biological Redundancy: Another important feature of nature that is being applied to robotics is the concept of redundancy and fault tolerance in systems. Biological systems are often able to adapt and survive damage, even when smaller biological systems fail. Similarly, How to implement a robotic design is a very important consideration. Exactly what methods and materials that will be used can determine the success or failure of a robot. The actuators and systems in place used to mimic actual biological mechanisms should, much like a real biological creature, be adapted to suit the environment and stress the robot will face. Examined are a few locomotion mechanisms. Shape Memory Effects: One type of biomimetic material that closely resembles muscles and is applied as an actuator in robotics is Nitinol. This material is a nickel-titanium alloy that behaves will contract and expand based on temperature applied, much as a muscle would when stimulated using electricity. Called the shape memory effect (SME), this interesting effect allows a wide variety of approaches in robotic biomimetics.[3][1] Pneumatic Actuators: . 2 Another muscle mimicry performed by robotics is done by using pneumatic actuators. This will mimic muscles by pressurizing rubber tubes with air which will cause them to expand and contract similarly to biological muscles. While this is an excellent lightweight alternative to some systems, the pressures and frictions that the tubes undergo often leave these types of muscles with a relatively short lifetime.[2][4][7] Hydraulic Actuators: Similar to the pneumatic actuators, hydraulic ones use fluid instead of gas. These allow higher loads to be used and are generally more durable. The drawback however, is that the systems needed to use hydraulics, such as the pumps, are generally large and cumbersome. This leads to a reduced use in robotic motion.[2] Electroactive Polymer: These materials are polymers that change shape and behavior when an electric current is passed through them. Due to the extremely wide variety of these materials, they have an equally wide application in robotics. The primary research focus in the field of robotics at this time is their use as biomimetic muscles, to be used to replace less adaptive systems.[1][2][5] Chemical Muscles: While not actually a muscle, the reciprocating chemical muscle is a small self-contained noncombustible system that produces large quantities of energy relative to its size, as well as actuator motion. Many robot designers may be more familiar with battery powered robots, chemical muscles provide an interesting alternative that has a variety of applications. Evolution: One of the biggest problems faced with robotics is the limitation of the materials and mechanisms, examined above. Approaching robotics through the eyes of evolution, new solutions to difficult problems can be procured. As a robot is built of a few relatively simple components such as sensors and actuators, co-evolution of the hardware and software can FIG. 1 : Robotic Lobster achieve impressive results. By simulating a robot using software and genetic algorithms, solutions that are typically difficult to obtain, such as a gait for a robot over a specific terrain or environment, can be found. This allows the design of a robot to go from having a human hand-design the robot through repeated trial and error, which could still lead to an inefficient design, to designing a robot by allowing it to organically adapt to the environment through which it will traverse. Examples: The following robotic examples employ one or several of the above mentioned biomimetic approaches. Though some of the evolutionary robotic projects might focus on simulations on the computer, all the examples were selected according to the requirement that they include a functioning hardware device 1. Robotic Lobster : The goal was to create an underwater robot imitating the Homarus Americanus. It focuses on remodeling the control signals whose timing has improved over the years depending upon a creatures response to animal muscles. Thus the need for use of actuators which match their biological models arised. SMA Nitinol turned out to be a best fit here considering its stress-strain relationship, relaxation and contraction speeds. Using it underwater with adequate insulation enhances its relaxation speed. Three degrees of freedom are provided to each leg of this eight legged robot. They are actuated by a pair antagonistic Nitinol wire which allows the leg to rotate bidirectionally. Elevation, swing and Side way motion are other few properties of this robot. To keep a check on Nitinols durability, the control signals are activated by PWM and have sufficient relaxation time. The control architecture consists of CPG network consisting of oscillating, command and coordinating neurons. The control 3 FIG. 2: Lamprey-based undulatory vehicle system chooses different behaviors according to sensor-input and supervised motivation of the robot 2. Lamprey-based undulatory vehicle This too uses Nitnol as actuator and Lampreys swim in rhythmic wave motion along the body axis. The efficient and highly pliable swimming locomotion of the lamprey makes it a choice for robotic emulation. It has a rigid hull which holds the processor, a plastic notochord which is highly flexible and a passive tail segment. Activation of the SMA wires allows axial bending movements around 5 joints. A CPG control architecture based on neural studies and outputs alternating PWM pattern signal is used. Due to its artificial muscles this robot produces strong autonomous forward swimming 3. SRI robots FIG. 3: CWRU’s Robot V 5. Cricket microrobot This type of robot designed after a cricket can sit in a 5cm cube and can walk and jump. Braided pneumatic actuators alon with spring bais are used as actuators for their joints. The actuator system consists of 16 air muscles and 32 valves. It employs a neural network control architecture. As a closedloop controller it includes the feedback from the angle sensors to adjust the activation patterns. SRI International’s six legged is loosely inspired by the cockroach. Each leg having 2 degree of freedom controlled by bow-tie actuators plus opposing spring forces. Its flapping robot emulates the flying insects. Their body consists of 4 Si bow-tie actuators which respond to their own resonance. SRI’s inchworm robot is a 16mm rolled actuator made from dielectric elastomer. Electrostatic clamps allows this robot to move in inchworm style on both vertical and horizontal surfaces. 4. Robot III + Robot V The close observation of running cockroaches were used in computer simulation and seven degrees of freedom were transferred onto each leg. All this was used to develop a new type of robot, robot III, a 75 cm long hexapod with aluminum tube legs that were modeled closely after the cockroaches morphology. The degree os freedoms vary for front, middle and rear legs. The total 24 degrees of freedom of the robots legs are actuated by double-acting pneumatic cylinders. Its control aarchitecture acts at posture, swing and stance level. . Due to pneumatic cylinders disability to deal load changes this robot could not create a steady walking pattern.[7]. The Robot IV replaced the pneumatic cylinder by pneumatic muscle. Its drawback is it force-to weight ratio. It fails to carry its own weight.[7]. As its predecessors robot V (also called Ajax) is non-autonomous and relies on external power supply and control. Its air muscle permits bidirection motion. FIG. 4: CWRU’s Cricket microrobot 6. Sprawl The ability of a cockroach to cross uneven terrain at a very afst speed while climbing obstacles without losing its speed makes it an ideal example for robot locomotion emulation in such varied conditions. It employs the pneumatic pistons as main actuators and servomotors for rotation of joints. Isprawl the youngest member in the family was without any actuators to allow autonomy. The use of polymers with different compliant values allows components to reach elasticity characteristics closer to their biological counterparts. This reduces the need for complex control signals to mimic these properties in software. 4 FIG. 4: This type of robot is excited by a pair of piezo-electric actuators which are at same voltage but 90 degree out of phase. This phase shifts brings in the two degrees of freedom required on each leg in parallel aiding elliptical foot motion. The drawback being such types of robot demand high voltage actuation. The robot is able to turn by a radius of 5cm when the excitation frequency is shifted [1]. Locomotion through vibration at a specific resonance frequency is a different approach for robotics. FIG. 5: Standford’s iSprawl 9. Snake robots 7. Scorpion As the name suggests this robot ahs eight legs having three joints actuated by DC motors and allow protraction/retraction. With the use of DC motors each leg of this robot can lift 8 times its own weight The control system is based on the CPG model and is distributed to one global and several local computing units. Use of oscillators yield rhythmic motion patterns. This also employs reflex control mechanism which alongwith CPG should be able to overcome uneven terrains. Effective limless motion strategies can be observed from a snake. S0 was constructed using 12 segments connected with servometers. But its working turned out to be disappointing due to fauty weight distribution and energy loss due to friction. S3 is equipped with two servomotors and two degree of freedoms. Robot S5 has 32 joints. The latest one is without the wheels and employs sonar and heat sensors and a compass. FIG. 8: Snake robot S7 FIG. 6: Fraunhofer Institute’s Scorpion 8. Mesoscale robot quadruped 10. RoboTuna RoboTuna is modeled after Bluefin Tuna. Its first generation was 1.25m long underwater robot having aluminium skeleton and a hull ribs covered in lyrca skin The movement of the backbone is through sic servo meters who energize it through cable and pulley system. FIG. 7: Mesoscale robot quadruped FIG. 9: RoboTuna 5 11. VCUUV VCUUV stands for The Vorticity Control Unmanned Undersea Vehicle. As opposed to the RoboTuna, VCUUV is fully autonomous. It consists of an energy source, hydraulic actuator and a control unit. . Stabilized by the caudal fin the VCUUV propels and maneuvers steady in the horizontal plane. Tests of the maneuverability of the vehicle show that its performance matches those of real tunas [1][6]. FIG. 12: Neural oscillator control network for Rodney 14. OCT-1b The octopod robot OCT-1b is designed after the body of a lobster. Its eight metal legs are each actuated by two servomotors that cause them to lift and swing. The robot was used for different walking pattern evolutions that were carried out in 1998. FIG. 10: VCUUV 12. Entomopter It tries to ape locomotion strategies of crawling, flying and maneuvering while flying insects. It uses an ultrasonic range finder for navigation which aids autonomous navigation. FIG. 13: Oct-1b 15. SONY quadruped robot 13. Rodney Four legged robots each having three degrees of freedom created by Sony Corporation installs infrared sensors and a camera for fast color detection. Preprogrammed procedures help the robot to stand up and reposition itself and start from its nitial position . Rodney is ahexapod whose each leg is actuated y two servomtors to produce limb lift and swing. The control architecture for the servomotors of the robot was designed as a simple neural oscillator network withweighted connections. The output of neural network is then downloaded into the hexapod and its results verified by human researchers. FIG. 14: Sony quadruped robot 16. Koharo - Crawling and Jumping Deformable Soft 6 Robot the point of error and creates a new best locomotion behavior to compensate for the physical damage [10]. The Koharo project uses deformation as an alternative approach for locomotion over rough terrain. By contracting the additional coils elastic potential energy gets stored and if released fast enough makes the robot jump a distance twice its diameter [8]. FIG. 17: Self-modeling robot 19. Golem FIG. 15: Koharo The Genetically Organized Lifelike Electro Mechanics Golem uses the simple electromechanical systems with autonomous physical construction using rapid prototyping technologies. Bars, actuators and artificial neurons are the elementary building blocks for its design. It also employs microcontrollers and neural networks 17. Random morphology robots The RM-robots has 6 servo-motors linked together randomly by thin metal joints. Its seen that as the control architecture grasps the uneven morphology and generates locomotion patterns. The analysis of the best control programs shows that faster forward locomotion is usually reached by a more complex and therefore longer program [9]. FIG. 18: Golem 20. Genobots The building blocks for the Genobots are simple bars connected through fixed or actuated joints that have a moving angle of 60 degrees. Frequency and phase offset of the actuators are parameters of the form finding process and ensure the co-development of morphology and control architecture. FIG. 16: Random morphology robot 18. Self-modeling robot This system studies its own morphology from scratch and updates itself on unexpected damage to its structure. During tests, the systems reads the observed and the predicted behavior and restartrs a model finding process. After several cycles of testing and remodeling the system is able to discover FIG. 19: Genobots 7 [6] Anderson, J. M. & Chhabra, N. K. (2002) ’Maneuvering and stability performance of a robotic tuna’, Integrative and Comparative Biology, vol. 42, pp. 118- 126. 21. Nonaped Its an attempt at evolving dynamic walking patterns for a physical robot. The robot has two stewart platforms and have 12 degrees of freedom. Each pneumatic actuator is mounted between two platforms with ball-and-socket joints. Individual actuation of the pneumatic piston allows the platforms pitch, roll and yaw freedom. [7] Kingsley, D. A., Quinn, R. D. & Ritzmann, R. E. (2003) ’A cockroach inspired robot with artificial muscles’, Proceedings of the International Symposium on Adaptive Motion of Animals and Machines, Kyoto, Japan. [8] Shiotsu, A., Yamanaka, M., Matsuyama, Y., Nakanishi, H., Hara, Y., Tsuboi, T., Iwade, T., Sugiyama, Y. & Hirai, S. (2005) ’Crawling and jumping soft robot KOHARO’, Proceedings of 36th International Symposium on Robotics, Tokyo [9] Dittrich, P., B¨urgel, A. & Banzhaf, W. (1998) ’Learning to move a robot with random morphology’ in Evolutionary Robotics - First European Workshop, ed. P. Husband & J-A. Meyer, Springer Verlag, Heidelberg. [10] Bongard, J., Zykov, V. & Lipson, H. (2006), ’Resilient machines through continuous self-modeling-, Science, vol. 314, no. 5802, pp. 1118 - 1121. [11] Hornby, G. S., Lipson, H. & Pollack, J. B. (2001) ’Evolution of generative design systems for modular physical robots’, IEEE International Conference on Robotics and Automation, Piscataway, NJ. FIG. 20: CCSLs Nonaped [12] Clark, J. E., Cham, J. G., Bailey, S. A., Froehlich, E. M., Nahata, P. K., Full, R. J. & Cutkosky, M. R. (2001) ’Biomimetic design and fabrication of a hexapedal running robot’, International Conference on Robotics and Automation, Seoul, Korea. III. CONCLUSION While animals well-suited to their environments and perfectly capable adapting to and learning new things, robots are largely per-determined and incapable of these same capabilities. The aims of biomimetics are to give robots the same capability and adaptability, even survivability, of their animal counterparts. The ability to repair itself, learn from mistakes, adapt to suit the environment, and even think are all goals to biomimicry. With the use of more creative materials and ideas, robots will soon be just as capable as animals, if not even better in many regards. REFERENCES [1] Ayers, J., Davis, J. L. & Rudolph, A. (2002) Neurotechnology for biomimetic robots, MIT Press, Cambrdge [2] Bar-Cohen, Y. & Breazeal, C. (2003) Biologically inspired intelligent robots, SPIE PRESS, Bellingham,Washington [3] Gilbertson, R. G. (1993, 2000) Muscle wires project book, Mondotronics, San Rafael [4] Iovine, J. (2002) Robots, androids, and animatrons, McGraw-Hill, USA. [5] Ashley, S. (2003) ’Artificial muscles’, Scientific American, vol. 289, no. 4 , pp 5259 .