Reinforcement Learning applications for robots, stabilization of biped (passive dynamic walking) robot, and information security involving traceability of fresh food circulation.
Attractive Factors of My Research
My research interest is to make strong mechanical,robotic, and information systems by simple controll of essential part of systems. First example is a machine learning agent that gains co-ordinated behaviour through interaction without communication. This is like passengers at a crowded ticket gate, where individuals manage to go through their way without communicating, while some rules emerges in that area. The idea of prediction by dynamics reconstruction is used with Reinforcement Learning scheme. Some examples include traffic signal control and multi-legged robot recovery.
The other example is a passive dynamic walking biped robot that is capable of walking on an outdoor road. The passive dynamic walk is like a sequence of inverted pendulum’s sway and is known as an energy-efficient way but also is very condition-sensitive. My idea is to introduce a self-maintaining stability, like a ship having stability by its bottom shape, into a sole shape so as to produce self-recovery force. Our experimental robot has broken the previous Guinness World Record (13 hours) of non-stop walking by passive dynamic robot by 100 hours.
The last example is a control of food circulation system. We added a simple shipping volume recording system into a traceability system for Shijimi clam in Aomori prefecture. The system is currently in practical use and is used to protect against deceptive labeling of orgins of the clam.
- Publication: Publised a Japanese interpretation of a textbook of Reinforcement Learning (2000).
- Robot technology: Our biped walking machine was certified by the Guinness World Record of the longest walk by passive dynamic robot (16.3 km/100 hours) (2011).
- Practical applications: A pioneering work for a traceability system for Shijimi clam is commercially used in Aomori prefecture.
- Awards: “Planning Award for 2008 Technical Meetings of Division C,” The Institute of Electrical Engineers of Japan (2009),“Technical Research Paper Award,” Information Processing Society Japan Hokkaido Branch (2004),“Contribution Award of Robotics and Automation Division,” Japan Society for Mechanical Engineering (2004), “Theoretical Developments in Techniques Award,” Artificial Neural Networks in Engineering (US)(1997), Robotics and Mechatronics Paper Award,” Japan Society for Mechanical Engineering (1997)
Major Books and Papers
- S. Mikami, T. Fukuda, and S. Suzuki, “Gait Design by Graph Coloring for Robots That Have Legs on Their Faces,” in 2015 IEEE/ASME International Conference on Advanced Intelligent Mechatronics, Busan, Korea, 2015, p. WeAT4.3-1-5.
- T. Omido, S. Mikami, and K. Hyodo, “Limiting Stance Leg Heel Angle by Sole Shape to Achieve Stable Passive Dynamic Walk,” in 13th International Conference on Control, Automation, Robotics and Vision, Singapore, 2014, p. Th37.6.
- S. Mikami, T. Fukuda, A. Igarashi, and S. Suzuki, “Prismatic Omni-Directional Legged Machine for Rough Environment Movement by a Simple Mechanical Drive,” in 13th International Conference on Control, Automation, Robotics and Vision, Singapore, 2014, p. Fr43.4.
- S. Mikami, N. Rikiishi, M. Sugawara, K. Hyodo, “The longest walk by a passive walking robot.” Guinness World Records, Apr-2013.
- N. Rikiishi, M. Sugawara, T. Omido, K. Hyodo, and S. Mikami, “Walk Support Shoes by the Principle of the Passive Dynamic Biped Robot Stabilization Using Constraining Foot Shape,” Journal of Biomechanism Society Japan, vol. 37, no. 1, pp. 66-69, Jan. 2013. (in Japanese)
- T. Wakahara and S. Mikami, “Feedback Control of Traffic Signal Network of Less Traffic Sensors by Help of Machine Learning,” in 12th International Conference on Intelligent Autonomous System, Jeju Island, Korea, 2012, pp. T4C-S5, 1-5.
- T. Wakahara and S. Mikami, “Adaptive Nutrient Water Supply Control of Plant Factory System by Reinforcement Learning,” JACIII, vol. 15, no. 7, pp. 831-837, Sep. 2011.
- K. Shimbori, K. Hyodo, K. Sunayama, and S. Mikami, “Acquisition of Adaptive Movement Strategy by Reinforcement Learning in Modular Multiple-leg Mobile Robot,” Journal of Information Society Japan, vol. 50, no. 3, pp. 1170-1180, Mar. 2009.(in Japanese)
- K. Hyodo, S. Suzuki, and S. Mikami,“Designing a Foot Shape that Stabilizes Passive Dynanlic Walk by Constraining Leg Angles,” Journal of JSME, vol. 74, no. 742, pp. 1514-1521, 2008.(in Japanese)
- S. Mikami, Y. Takahashi, A. Nagano, S. Kuwabara, H. Narumi, and T. Wakabayashi, “Enabling Secured Traceability of Fishery Products Using 2D Code and Digital Encryption,” in 12th Biennial Conference of the International Institute of Fisheries Economics and Trade, Tokyo, Japan, 2004, pp. 168-1-7.