Constructive understanding of snakes' adaptive locomotion
Biological system models reproducing snakes' musculoskeletal system
Snakes exhibit very high environmental adaptability by bending their long cord-shaped body. In order to elucidate functional significance of their special body structures and adaptive control mechanisms specific to their body structures, we are constructing biological system models reproducing their musculoskeletal systems. We conducted biological analyses including dissection, CT-imaging and motion capture to design models as real as possible. Our models include the virtual snake simulator and the snake-like robot with very long pneumatically actuated artificial muscles (PAS-2). Using these models, we are aiming at elucidation of mechanisms underlying dextrous adaptivity of snakes.
Analysis of skeletal system of snakes with CT-imaging
Snake-like robot with hyper-articulated muscles, PAS-2
Lateral undulation locomotino by PAS-2 (movie)
Kousuke INOUE, Kaita NAKAMURA, Masatoshi SUZUKI, Yoshikazu MORI, Yasuhiro Fukuoka, Naoji SHIROMA: Biological System Models Reproducing Snakes' Musculoskeletal System, Proc. IEEE/RSJ International Conference on Intelligent Robots and Systems (IROS2010), pp.2383-2388 (2010/10)
CPG-based control of a snake-like robot
Many kinds of animals' locomotions including swimming, walking, meandering and fluttering are rhythmic and coordinated over whole body. What the recent researches elucidate is that this coordinated rhythm is not generated in upper centers, e.g. brain cortex, but is generated and controlled by lower centers called CPG (central pattern generator), which is distributed in spinal cord (vertebrates) or nerve ganglions (invertebrates). In this research, we constructed a CPG-based control system for lateral undulation locomotion of snakes according to the model that Ekeberg proposed for lamprey. By feeding back information of reaction force from the ground into CPG, we achieve a neural controller that enables a snake model to creep with smaller winding for higher friction and bigger winding for smaller friction, which is similar to adaptation in living snakes.
Snake like robot used as a model
CPG model (Ekeberg)
Experiment using robot (movie)
Kousuke INOUE, Shugen MA, Chenghua JIN: Neural Oscillator Network-Based Controller for Meandering Locomotion of Snake-Like Robots, Proceedings of 2004 IEEE International Conference on Robotics and Automation (ICRA2004), pp.5064-5069 (2004/04)
Kousuke INOUE, Shugen MA, Chenghua JIN: Optimization of CPG-Network for Decentralized Control of a Snake-Like Robot, Proceedings of 2005 IEEE International Conference on Robotics and Biomimetics (ROBIO2005), pp.730-735 (2005/07)
Kousuke INOUE, Takaaki SUMI, Shugen MA: CPG-based Control of a Simulated Snake-like Robot Adaptable to Changing Ground Friction, Proc. IEEE/RSJ International Conference on Intelligent Robots and Systems (IROS2007), pp.1957-1962 (2007/10)
Constructive understanding of Myriapoda adaptive locomotion
Analysis of distributed control of millipedes walking
Myriapoda have a large number of legs along with their long body and motion of the legs is highly coordinated and adaptive to surrounding environments. We investigate the distributed control mechanisms of the legs.
Tanzanian giant black millipede (Spirosteptus giganteus)
Emi SATO, Toshihiro TAMURA, Kousuke INOUE: Experimental Investigation of Walking Pattern Adjustment of Millipedes, Autonomous distributed systems, 2015 (in Japanese)
Iterative Transportation by Cooperative Mobile Robots
Iterative Transportation of many objects by multiple mobile robots is one important task category in applications of multiple robots. We proposed a motion planning method to realize this task in unknown or dynamic environments. The method includes cooperative exploration of the working environment, generation of path-network appropriate for the task, and formation construction to realize efficient cooperation in robots.
Iterative transportation task
Kousuke INOUE, Jun OTA, Tomokazu HIRANO, Daisuke KURABAYASHI, Tamio ARAI: Iterative Transportation by Cooperative Mobile Robots in Unknown Environment, Distributed Autonomous Robotic Systems 3 (Proceedings of The 4th International Symposium on Distributed Autonomous Robotic Systems (DARS1998)), Eds. Leuth,T., Dillmann,R., Dario,P., Worn,H., Springer, pp.3-12, (1998/05)
Kousuke INOUE, Jun OTA, Tamio ARAI: Iterative Transportation by Multiple Mobile Robots Considering Unknown Obstacles, Journal of Robotics and Mechatronics, 21-1, pp.44-56 (2009/02)
Behavior Acquisition by an Artificial Agent in Partially Observable Environment
Artificial agents operating in the real world suffer from problems in observation, such as limited sensing range, occulusion, and sensing error. Furthermore, the way of interpretation of specific sensor values is essentially dependent on the objective of the agent and the way to interact with the environment. Therefore, the way of interpretation is very difficult to predefine. In this research, a method to learn the way of recognition according to agent's experience to solve the above-mentioned problems by constructing internal environmental model containing the context based on POMDP (partially observable Markov decision process) concept.
Kousuke INOUE, Shugen MA, Jun OTA: Acquisition of Global Information from Local Observation with Movement - Construction of Internal State-Representation under Partial Observabability -, Proceedings of IEEE International Conference on Robotics, Intelligent Systems and Signal Processing (RISSP2003), pp.370-375, (2003/10)
Acquisition of Generalized Knowledge for Ontogenetic Development of a Robot
The time scales of adaptation of an artificial agent can be classified as follows: (1) Short-term adaptation: behavior acquisition specific to given environment and task, (2) Phylogenetic adaptation: adaptation of a species in generations by the use of evolutional methodologies, and (3) Ontogenetic adaptation: "development" of a single agent by experiencing various tasks and environments. In this research, we focus on adaptation in the third time-scale. We proposed a method to accumulating generalized knowledge through multiple tasks and environments by analysing generality of each experience.
Kousuke INOUE, Jun OTA, Tomohiko KATAYAMA, Tamio ARAI: Acceleration of Reinforcement Learning by a Mobile Robot using Generalized Rules, Proceedings of 2000 IEEE/RSJ International Conference on Intelligent Robots and Systems (IROS2000), pp.885-890, (2000/04)
Kousuke INOUE, Jun OTA, Tamio ARAI: Acceleration of Reinforcement Learning by a Mobile Robot Using Generalized Inhibition Rules, Journal of Robotics and Mechatronics, 22-1, pp.122-133, (2010/02)