On Coordinated Control Of Multiple Bio-mimetic Robotic Fish

In the future, the bio-mimetic robotic fish will be engaged in the tasks such as benthos observation, oceanic exploration, military reconnaissance and underwater construction in the complex underwater environment. Thus it is important to develop some efficient, flexible and robust coordination methods for a group of bio-mimetic robotic fishes. Due to the complexity of the underwater working environment and the particular propulsion mechanism of the robotic fish, the ordinary coordination approaches can not be applied to the robotic fish directly. As a result, coordinated control of multiple bio-mimetic robotic fish remains a challenging issue. Based on the successful development of the robotic fish platform, inspired by the social behaviors in nature and supported by the multi-robot coordination and control techniques, this thesis investigates the coordination methods of multiple bio-mimetic robotic fish.First, the research background of the bio-mimetic robotic fish and their coordination methods are introduced. In addition, the relative research work is synthesized, and the main result of this thesis is simply outlined.Second, based on the dynamic model of multi-joint bio-mimetic robotic fish, an optimization algorithm for our three-joint robotic fish on its straight tour performance is proposed using a general optimization method. In addition, the idea of fuzzy control is introduced in the velocity coordination control of multiple robotic fishes.The coordination control of two robotic fish is then studied in the environment with obstacles. The operation space of robotic fish is converted into a virtual force field by using the artificial potential field method. The force relationship models between the two robotic fishes, the robotic fishes and the target points, the robotic fishes and the obstacles are established. Each robotic fish can complete the task of swimming from the starting points, avoiding collisions with obstacles and the other robotic fish, and reach the target point successfully. This may provide some guidelines on the task implementation for robotic fishes in the complex environment in the future.In addition, an experiment of two robotic fishes passing through the hole in turn is designed. This experiment is based on the reinforcement learning algorithm. The robotic fishes attempt different movements in different states, then get the rewards according to the result. The next action is determined according to the rewards. A better action strategy in different states is realized for the autonomous robotic fishes.Finally, a conclusion is summarized and the future work is pointed out.