Clonal Selection Optimization Based Adaptive Control For Ship Steering

Based on ideas gleaned from the clone selection principle in immunology system, Clone Selection Algorithms (CSA) has become an advanced method, offering a fresh perspective for complicated calculation problems. CSA attracts foremost attention among calculation researchers. Accordingly, Ship Motion Control is a study focus attended by scientists of both navigation science and control theory because the performance of autopilot for ship steering is of great importance for a ship to navigate safely and economically. Aiming at the research of adaptive control for ship steering, this dissertation examines the Clone Selection Algorithms for global optimization and applications to the adaptive control for ship steering.1. A Chaos-Clone based Evolutionary Algorithm (CCEA) is proposed by integrating chaos search and clone selection algorithm (CLONALG). CCEA uses chaotic floating point numbers code instead of the binary code of CLONALG, and produces the initial diversity of antibody population. The algorithm adopts a chaotic disturbance strategy for the antibodies with high affinity; the different chaotic disturbance is added to an antibody according to its affinity to antigen; also, the disturbance factor changes with the evolutionary generation so as to speed search during prophase and convergence during anaphase. CCEA uses a chaos to reshuffle operation for those antibodies with low affinity to maintain the diversity of the population. Simulations on complex benchmark functions demonstrate that the CCEA has better performance than both the chaos optimization and CLONALG individually used.2. A dynamic immune clone selection algorithm with classified mutation is proposed based on traditional floating point coding. Applying grading mutation and dynamic parameter strategies, the method speeds up the global search and improves the local con- vergence precision in the process of inner population mutation and their evolution. On one hand, according to the antibody affinity in relation to the antigen, the antibody population is decomposed into three subsets, and they are submitted to respective mutation processes for their different given tasks. On the other hand, the population size, the clone size and the mutation parameters are dynamically changed with evolution process. The proposed algorithm is use to optimize complex functions for testing, and the results show merits of its effectiveness in CSA, such as high accuracy, quick convergence and less opportunity for local optima.3. Based on clone selection principle, a novel evolutionary algorithm is proposed with the aid of traditional method using elitism clone mutation and heuristic crossover. The two main operations of elitism clone mutation and heuristic crossover are defined. The elitist antibodies with highest affinity are subject to a small mutation process to search local optima. Those antibodies with general affinity are suffered to a heuristic crossover with elitist antibodies to speed global optimization. The antibodies with lowest affinity are replaced by new individuals to maintain the diversity of the population. Also, to prevent evolutionary stagnation, the scaling factors of affinity are adaptively adjusted in order to guarantee the accuracy of local convergence when speeding up global searches. The computer simulation results adopted complex benchmark functions demonstrate that the proposed algorithm has good performance in the aspect of improvement of CLONALG, easy operation, and high accuracy.4. A class of model-based clone selection adaptive control algorithm is developed by applying CSA to adaptive control problem. CSA dopes out the plant output by a plant model and evaluates the candidate controllers before selecting the fittest controller to the current plant operating conditions. Such approach realizes on-line controller tuning. Then, a CSA direct adaptive PID control is developed by applying this approach to conventional PID controller for ship steering. Moreover, we introduce two new, but closely related approaches to CSA adaptive control which is called "CSA reference model supervisory adaptive PID autopilot" and "CSA reference model tracking adaptive PID autopilot" for ship steering. In these techniques a reference model is used to guide the system output towards the reference model output. Simulations verified that CSA can carry on the adaptive control excellently, the above three adaptive control algorithm for ship steering have good robustness to environmental disturbance and ship model error.5. Two CSA model-identified model reference PD controllers, respectively with and without Disturbance Identification-and-Compensation (DIC) are proposed to resist both ship model possible uncertainty and changeful operating conditions. Regarding ship system as a "black box ", CSA on-line identifies the ship with a second order linear model according to the information of both course and rudder angle. Then, the second order linear together with a model reference is used to calculate the PD controller parameters. Their goals are to steer the ship tracking reference model output. Simulations show the controller with DIC has better performance than that without DIC, though the controller without DIC can also cope with the disturbance by translating some disturbance into plant model parameters.