| With the constant development of technology,an environmental protection issue is increasingly valued by different countries.Transportation is a key field for lowcarbon environmental protection,energy conservation and emission reduction.In recent years,with the occurrence of two-wheel balance cars in the market,due to simple structure,energy conservation and environmental protection,small radius of turning circle and convenient control,it becomes the tool for riding instead of walk for lots of travelers and also attracts attention of many scientific researchers.By aiming at the features of the balance cars’ control system,the author designs the balance cars’ control system and conducts relevant studies for the balance cars’ control algorithm.The main research contents and achievements of the thesis are stated as follows:1.By aiming at the kinetic characteristics and structure features of two-wheel balance cars,the balance cars’ kinetic equation is constructed through the Lagrange equation.Moreover,MATLAB is used for simulation to verify correctness of this model.2.Based on the two-wheel balance cars’ control system,the overall scheme of control is designed.Also,the author selects the corresponding hardware module for the control design of the hardware circuit.Among which,the hardware module gets involved in the STM32 SCM module,MPU6050 sensor module,power module and motor drive module.Software mainly uses the complementary filtering algorithm to solve the balance cars’ posture.In the meantime,PID algorithm is used to control the balance cars.At last,MDK software burns the program to the STM32 SCM for controlling,so as to realize self-balancing.3.Reinforcement learning control algorithm is utilized to do simulation control for balance cars.In fact,this algorithm dispenses with a model and priori knowledge.By constantly conducting the trial and error learning,the control strategy is exercised.In the simulation experiment,an initial state is given to balance cars.The experimental effect reaches the expected result,proving the effectiveness of the reinforcement learning algorithm.4.Targeting at the malpractice of the pure reinforcement learning control,the author designs the algorithm by combining with reinforcement learning with the neural network.The improved reinforcement learning algorithm is adopted to do simulation control for balance cars.By comparing with the pure reinforcement learning algorithm,the simulation experiment effect is even more ideal. |
