Parameter optimisation of path‐following fuzzy controller for a six‐wheel lunar rover

Abstract

Path following is a main and fundamental task for future lunar rover autonomous navigation. This study adopts genetic algorithm (GA) to optimise the parameters of a path-following fuzzy controller designed for a six-wheel lunar rover. Considering the influences of the orientation deviation and its variation rate to the controller performance, two quantisation factors and one scale factor are utilised to limit both the input and output variables. However, it is difficult to manually achieve the proper factors because they are coupling themselves complicatedly. They are adjusted and modified automatically during the path-following process based on GA to achieve the auto-tradeoff the parameter and its control performance. Those parameters are coded using floating-point encoding scheme, respectively. The minimum tracking error and the angular velocity are taken as the target function. After several iterations of crossover and mutation, the best parameters are determined aiming at achieving lower path tracking error with small angular velocity. Simulation on different paths and comparisons with traditional fuzzy controller results show the designed controller with optimised parameters has better performance than the one with manually regulated parameters.

Design of a hybrid controller for voice coil motors with simple self-learning fuzzy control

“…Thus, the SSFC can automatically tune the fuzzy control rules base to achieve satisfactory performance. (15) where s is the Laplace operator, ξ and n w are the damping ratio and undamped natural frequency.…”

“…The fuzzy rules should be pre-constructed to achieve the design performance by trial-and-error; however, this trial-and-error tuning procedure is time-consuming. To overcome the trailand-error tuning of the membership functions and fuzzy rules, the fuzzy control scheme has been combined with many different methods to tune the fuzzy control rules [10][11][12][13][14][15][16]. In [10][11][12], the adaptive fuzzy control approach is designed to online tune the fuzzy rules in the Lyapunov stability theory; however, the approximation error between the system uncertainty and fuzzy uncertainty observer may cause instability of the closed-loop system.…”

“…In [13,14], the fuzzy neural network approach with parameter learning are proposed by using backpropagation learning algorithm, but it is based on gradient descents that is easily trapped at local minima. In [15,16], the evolution algorithm has been successfully applied to solve many optimization problems. However, it always leads to heavy computational costs and the convergence speed may be slow.…”

Self-organizing fuzzy sliding-mode control for a voice coil motor

“…To resolve this problem, several learning methods have been proposed [8][9][10]. Though favorable control performance can be achieved in [8][9][10], the learning algorithms only take care of parameter learning but neglect structure learning of fuzzy rules.…”

“…Though favorable control performance can be achieved in [8][9][10], the learning algorithms only take care of parameter learning but neglect structure learning of fuzzy rules. Time-consuming trial-and-error process is needed to determine the number of fuzzy rules.…”