Abstract
In this paper, we propose two methods of adaptive actor-critic architectures to solve control problems of nonlinear systems. One method uses two actual states at time k and time k+1 to update the learning algorithm. The basic idea of this method is that the agent can directly take some knowledge from the environment to improve its knowledge. The other method only uses the state at time k to update the algorithm. This method is called, learning from prediction (or simulated experience). Both methods include one or two predictive models, which are assumed to be applied to construct predictive states and a model-based actor (MBA). Here, the MBA as an actor can be viewed as a network where the connection weights are the elements of the feedback gain matrix. In the critic part, two value-functions are realized as a pure static mapping, which can be reduced to a nonlinear current estimator by using the radial basis function neural networks (RBFNNs). Simulation results obtained for a dynamical model of nonholonomic mobile robots with two independent driving wheels are presented. They show the effectiveness of the proposed approaches for the trajectory tracking control problem.
Similar content being viewed by others
References
Sutton RS, Barto AG (1999) Reinforcement learning, an introduction. MIT Press, Cambridge
van Buijtenen WM, Schram G, Babuška R, Verbruggen HB (1998) Adaptive fuzzy control of satellite attitude by reinforcement learning. IEEE Trans Fuzzy Syst 6(2):185–194
Berenji HR, Khedkar P (1992) Learning and tuning fuzzy logic controllers through reinforcements. IEEE Trans Neural Netw 3(5):724–740
Barto AG, Sutton RS, Anderson W (1983) Neuronlike adaptive elements can solve difficult learning control problems. IEEE Trans Syst Man Cybern 13(5):834–846
Juang CF, Lin JY, Lin CT (2000) Genetic reinforcement learning through symbiotic evolution for fuzzy controller design. IEEE Trans Syst Man Cybern B 30(2):290–302
Kandadai RM, Tien JM (1997) A knowledge-base generating hierarchical fuzzy-neural control. IEEE Trans Neural Netw 8(6):1531–1541
Chiang CK, Chung HY, Lin JJ (1997) A self-learning fuzzy logic controller using genetic algorithms with reinforcements. IEEE Trans Fuzzy Syst 5(3):460–467
Doya K (2000) Reinforcement learning in continuous time and space. Neural Comput 12(1):219–245
Watanabe K (1992) Adaptive estimation and control. Prentice Hall, London
Watanabe K, Syam R, Izumi K, Kiguchi K (2001) Adaptive actor-critic designs with current estimated or predicted value-function. In: Proceedings of the 5th international conference on KES2001, Osaka, pp 1308–1318
Prokhorov DV, Wunch DC (1997) Adaptive critic designs. IEEE Trans Neural Netw 8(5):997–1007
Fierro R, Lewis FL (1998) Control of nonholonomic mobile robot using neural networks. IEEE Trans Neural Netw 9(4):589–600
Watanabe K, Tang J, Nakamura M, Koga S, Fukuda T (1996) A fuzzy-Gaussian neural network and its application to mobile robot. IEEE Trans Control Syst Technol 4(2):193–199
Syam R, Watanabe K, Izumi K, Kiguchi K (2001a) Adaptive actor-critic design using predictive model and its application to nonholonomic mobile robot. In: Proceedings of the 5th international conferenec on KES2001, Osaka, pp 1319–1324
Syam R, Watanabe K, Izumi K, Kiguchi K (2001b) Adaptive actor-critic learning of mobile robot using actual and simulated experiences. In: Proceedings of the international conference on ICCAS2001, Cheju, pp 312–316
Syam R, Watanabe K, Izumin K, Kiguchi K (2002) Control of nonholonomic mobile robot by an adaptive actor-critic method with simulated experience based value-functions. In: Proceedings of conference on IEEE-ICRA2002, Washington, pp 3960–3965
Author information
Authors and Affiliations
Corresponding author
Rights and permissions
About this article
Cite this article
Syam, R., Watanabe, K. & Izumi, K. Adaptive actor-critic learning for the control of mobile robots by applying predictive models. Soft Comput 9, 835–845 (2005). https://doi.org/10.1007/s00500-004-0424-1
Published:
Issue Date:
DOI: https://doi.org/10.1007/s00500-004-0424-1