Automatic Car Parking: A Reinforcement Learning Approach
The automatic parking of a car-like robot is the problem considered in this paper to evaluate the role played by formal representations and models in neural-based controllers. First, a model-free control scheme is introduced. The respective control actions are sensory-based and consist of a dynamic, neural-based process in which the neurocontroller optimizes ad hoc performance functions. Afterwards, a model-based neurocontroller that builds without supervised a formal representation of its interaction with the environment is proposed. The resulting model is eventually utilized to generate the control actions. Simulated experimentation has shown that there is an improvement in robot behavior when a model is used, at the cost of higher complexity and computational load.
Unable to display preview. Download preview PDF.
- 2.Paromtchik, I.E., Laugier, C. (1996) Motion generation and control for parking an autonomous vehicle. Proc. IEEE Int. Conference on Robotics and Automation, Minneapolis, MN, 3117–3122.Google Scholar
- 4.Kong, S.G., Kosko, B. (1992) “Comparison of fuzzy and neural track backer-upper control systems”. In B. Kosko (ed). Neural Networks and Fuzzy Systems. Prentice-Hall. Englewood Cliffs, NJ, 339–361.Google Scholar
- 7.Canudas de Wit, C. (1998) “Trends in mobile robots and vehicle control”. In B. Siciliano, K. P. Valavanis (eds). Control Problems in Robotics and Automation. LNCIS 230, Springer, Berlin, 151–176.Google Scholar
- 8.Maravall, D., de Lope, J. (2002) “A reinforcement learning method for dynamic obstacle avoidance in robotic mechanisms”. In D. Ruan, P. D’hondt, E. E. Kerre (eds). Computational Intelligent Systems. World Scientific, Singapore, 485–494.Google Scholar
- 9.Maravall, D., de Lope, J. (2003) “A bio-inspired robotic mechanism for autonomous locomotion in unconventional environments”. In C. Zhou, D. Maravall, D. Ruan (eds). Autonomous Robotic Systems: Soft Computing and Hard Computing Methodologies and Applications. Physica-Verlag, Springer, Heidelberg, 263–292.Google Scholar