Bionic Autonomous Learning Mechanism Study Based on Automaton and Applied on Robot

Abstract:

Under the learning automaton frame a bionic self-learning automaton which has autonomous learning mechanism is constructed based on conditioned reflex theory. The bionic self-learning automaton is a 5-tupe and mainly contains: 1) inner set, mainly are perception state set and operant set; 2) oriented core, guide the direction of autonomous learning; 3) learning core, the primitives of learning unit; 4) autonomous learning unit, implement learning computation; 5) operant entropy, measuring the self-organized degree of bionic self-learning automaton. Theory analysis is made for the designed bionic self-learning automaton in the thesis, which theoretically proves the convergence of autonomous learning mechanism in bionic self-learning automaton, namely the operant entropy will converge to minimum with the learning process. And then the designed bionic self-learning automaton which used as the mathematical abstraction and formal tool of robot and used to describe the self-learning behavior of robot is applied to motion control of two-wheeled self-balanced robot. The state sets of bionic self-learning automaton is described by tilt angle and tilt angle rate of robot, and the operant sets is described by control signal of left motor and right motor. Robot doesn’t have t motion balanced skill in initial state, and with the learning proceeding the operant entropy will gradually decrease. After about twenty-five rounds training, the operant entropy gradually tends to minimum. So robot gradually learned the motion balanced skill.

• Periodical: Applied Mechanics and Materials (Volumes 58-60)
• Pages: 1408-1414
• DOI: https://doi.org/10.4028/www.scientific.net/AMM.58-60.1408
• Online since: June 2011
• Authors: Jian Xian Cai, Hong Rui Ma, Quan Sheng Wang
• Keywords: Autonomous Learning, Conditioned Reflex, Learning Automaton, Motion Control, Operant Entropy

References

[1] R. A. Brooks: From Earwigs to Humans (Robotics and Autonomous Systems Publications, Netherlands 1997).

[2] R. Chatila: Robot mapping: An introduction (Springer Publications, German 2008).

[3] J. Lacroix: Toward a Visual Cognitive System Using Active Top-down Saccadic Control (International Journal of Humanoid Robotics Publications, Singapore 2008).

DOI: 10.1142/s0219843608001443

[4] D. Pierce: Learning to Explore and Build Maps (AAAI Publications, CA 1994).

[5] T. Dean: Inferring Finite Automata with Stochastic Output Functions and an Application to Map Learning (Machine Learning Publications, Netherlands 1995).

DOI: 10.1007/bf00993822

[6] A. EI-Osery: A Stochastic Learning Automaton Based Autonomous Control of Robotic Agents, (Institute of Electrical and Electronics Engineers Inc Publications, Tunisia 2002).

DOI: 10.1109/icsmc.2002.1167955

[7] X. Wang: Optimal Control of Robot Behavior Using Language Measure, (International Journal of Vehicle Autonomous Systems Publications, PA 2004).

[8] A. Cattinelli: Interacting with an Artificial Partner: Modeling the Role of Emotional Aspects, (Biological Cybernetics Publications, Germany 2008).

[9] R. Pfeifer: Self-Organization, Embodiment, and Biologically Inspired Robotics (Science Publications, Washington 2007).

[10] B. F. Skinner, in: The behavior of organisms, edited by Appleton-Century-Crofts, New York (1938).