Adaptive Tracking Control for Robots With an Interneural Computing Scheme

• Published in: IEEE transaction on neural networks and learning systems Vol. 29; no. 4; pp. 832 - 844
• Main Authors: Tsai, Feng-Sheng, Hsu, Sheng-Yi, Shih, Mau-Hsiang
• Format: Journal Article
• Language: English
• Published: United States IEEE 01.04.2018; The Institute of Electrical and Electronics Engineers, Inc. (IEEE)
• Subjects: Actuators; Adaptation models; Adaptive control; Adaptive tracking; behavior-based navigation; Biological neural networks; Computation; Computational modeling; Computer simulation; Energy conservation; Energy consumption; evolutionary neural networks; Fault tolerance; flow elimination; interneural computing; Mobile robots; Navigation; Navigation behavior; Neural networks; neural path pruning; nonlinear dynamics; Optimization; pattern learning; Pruning; Punishment; Robot control; Robots; Target tracking; Tracking control; Tracking problem; Trajectory analysis; Uncertainty; Visualization
• ISSN: 2162-237X; 2162-2388; 2162-2388
• DOI: 10.1109/TNNLS.2017.2647819

Adaptive tracking control of mobile robots requires the ability to follow a trajectory generated by a moving target. The conventional analysis of adaptive tracking uses energy minimization to study the convergence and robustness of the tracking error when the mobile robot follows a desired trajectory. However, in the case that the moving target generates trajectories with uncertainties, a common Lyapunov-like function for energy minimization may be extremely difficult to determine. Here, to solve the adaptive tracking problem with uncertainties, we wish to implement an interneural computing scheme in the design of a mobile robot for behavior-based navigation. The behavior-based navigation adopts an adaptive plan of behavior patterns learning from the uncertainties of the environment. The characteristic feature of the interneural computing scheme is the use of neural path pruning with rewards and punishment interacting with the environment. On this basis, the mobile robot can be exploited to change its coupling weights in paths of neural connections systematically, which can then inhibit or enhance the effect of flow elimination in the dynamics of the evolutionary neural network. Such dynamical flow translation ultimately leads to robust sensory-to-motor transformations adapting to the uncertainties of the environment. A simulation result shows that the mobile robot with the interneural computing scheme can perform fault-tolerant behavior of tracking by maintaining suitable behavior patterns at high frequency levels.