Research On Motion And Control Embedded System Of Spherical Underwater Robot

Spherical underwater robot, the precise kinetic equations of which are usually hard to acquire, is a nonlinear system with strong coupling motions of each freedom. The motion control is the basis of the development of the spherical underwater robot system, and the motion control system is the precondition for the spherical underwater robot to finish the tasks and missions. This paper aims at the application of neural network control techniques to motion control of spherical underwater robot, and trying to design a control system with nice performance for a certain spherical underwater robot.Firstly, we do research on BYSQ-2 spherical underwater robot, and analyze the spherical underwater robot system in detail, put forward a method which bases on the control algorithm of PID neural network, and use the method to simulate this system. Secondly, this thesis designs a set of embedded motion control system based on AT91SAM7X256. What's more, it briefly has an introduction for method of running and debugging of the controller.The spherical underwater robot is a complex control objects which have characteristics of great inertia, delay, and time-varying. Therefore, it is very difficult to obtain good control with the rules of regular cascade PID control. On the contrary, using the strategy of PID neural network (PIDNN) can obtain satisfactory results. The results of simulation shows that, comparing with conventional PID control, the PIDNN control not only improves the dynamic performance of the system, but also, to a certain extent, reduces the impact which the object have induce for its time-varying characteristics on the effectiveness of control.This paper describes hardware components and working principle of the control system in the motion control system. ARM7 AT91SAM7X256 is used as the processor. Accelerometer and gyroscope can respectively feedback the attitude and heading of the spherical underwater robot. The system uses them to form a close loop control system which be adjusted by the neural network algorithm.In the part of software design, the basic software design technology and program design process of ARM are introduced. The C programming language is adopted in ARM. We introduce the PIDNN algorithm, online learning theory, data receiving of sensors and communication program design in detail. The paper use VC6.0 as a PC development environment to design the interface. VC MSComm control component is used in serial port interface and asynchronous communication is used between the PC and the ARM.