Attitude Control System Of The Two-wheel Self-balanced Robot Based On Learning

Two-wheeled self-balancing robot (TSR) is a kind of essentially instable wheeled mobile robot. Its dynamics system is non-linear, multi-variable, serious coupling, uncertain parameters etc. TSR already becomes an imagine platform to verify several kinds of control theory. It depends on two parallel wheels to support the bodywork. The battery provides power, and two motors drive the vehicle moving. Attitude sensing system, attitude control algorithm and machine of the bodywork cooperate to control the system's balance. It has simple structure, moves flexiblely, suits for working at some narrow and dangerous spaces. TSR could perform complicate motion and manipulation tasks wh;ch the multi-wheeled robot could not do, and it is very adaptable to the great-change environment or complicated tasks, such as space exploration, battlefield scout, dangerous goods transportation etc., it may also be used in toy, education and service robo(?) fields. So the research on TSR system has some significant theory and reality meanings.In this paper, the current developing status of two-wheled self-balancing robot were summarized and summed up. According to the supposition ideal condition and the analysis of TSR's moving rules, Its dynamic model was developed by Lagrange Equation. Then its most greatly controllable angle was inferred. All of the above job provided a theoretical basis for the design of TSR hardware and software system.This paper put forward planning of construction of the two-wheel self-balanced robot and design two-wheel self-balanced robot's mechanism.In order to monitor the motion pose and provide dependable data for the tracking control of the robot, TSR attitude measure system was designed with a micro-silicon accelerometer and a micro-silicon gyroscope. It provides the TSR's inverse angle and angle velocity to monitor the TSR's motion attitude. TSR attitude control system was constructed by FPGA.The micro-silicon accelerometer and the micro-silicon gyroscope are the main sensors of TSR attitude measure system, their precision immediate influence overall system precision and performance. The micro-silicon inertia component is more prominent in the volume and the cost merit, but has the very big insufficiency in the resolution and the precision, comparing with the micro-silicon accelerometer, the micro-silicon gyroscope's development lags. In order to abate the disadvantage effect caused by the errors of the accelerometer and the gyroscope, it has carried on calibration of the micro-silicon accelerometer and the micro-silicon gyroscope, based on this has used Kalman filter algorithms carrying on the fusion to the micro-silicon accelerometer and the micro-silicon gyroscope information, calculated tilt angle of the robot and has carried on the comparison to each algorithm, has obtained the Kalman filtering algorithm in the speed and the precision superiority. the attitude signal precision was improved. Because all of the TSR's locomotion styles are based on the balance control.The common control law is mainly of PID control,which is simple,effective and applicable.But the most obvious defect is the PID parameters can not be adjusted, the controller parameters can not be changed correspondingly while the working condition is changed, which will result to the decrease of control effect. Therefore, PID controller by using cerebellar model articulation control method which being adopted in fin stabilizer control is used in this thesis. Namely, when the working condition is changed, the PID controller by using cerebellar model articulation controller parameter adjusted automatically on line. In recent years, Reinforcement learning (RL), as a kind of machine learning algorithm, has a greatly development. RL doesn't need priori-knowledge and evolves its control behavior project by communicating with the environment. RL. Due to its self-learning characteristic, RL is applied to many fields, the representative example is the successfully controlling on the inverted pendulum. At the attitude control principle, TSR is the same as the inverted pendulum. Their centers of mass are in the upper portion of the body. And they keep balance by moving the lower part. So the design of TSR attitude control system was managed by the theory of RL to achieve balance control of the vehicle. Two-wheeled self-balancing electric vehicle means that all movements are to balance control as a precondition. Balance control is the key of its movement.In the end, the simulation and hardware experiments were carried out. Such as PID and RL simulation contrast experiment,The self-balanced experiment, the linear motion experiment and the anti-jamming experiment to the robot. The experimental result indicated that this system's performance can meet the design requirements and the design is rational.