A New Solving Method for Non-Linear Optimal Control Problem and Its Application to Automatic Berthing Problem
Naoki Mizuno, T. Kita, Takao Ishikawa · 2018
There are many researches for solving optimal control problems in control engineering, but the obtained input pattern is far differ from that of experienced operator or difficult to implement for real actuator. For example, the minimum-time optimal control solution becomes bang-bang type shape which cannot be implemented by slow actuator. In this paper, we consider a new solving method for optimal control problem with input pattern assignment using genetic algorithm and simulation method. Genetic algorithm is one of the evolutional algorithm to obtain numerical solution for optimization problem. However, the computational time using normal CPU is not sufficiently short for large scale genetic algorithm in which forward simulations of system dynamics are performed during sampling period. For this problem, the high performance GPU in recent years makes it possible to perform parallel genetic algorithm and simulation in very short time. The proposed algorithm is successfully applied to real system.