Two-wheel self-balance vehicle's machine learning control method

The invention discloses a two-wheel self-balance vehicle's machine learning control method, comprising: s1) obtaining the current state amount of the balance vehicle wherein the state amount includes the inclined angle and the angular velocity; s2) according to the probability for the optimal behavior, the state behavior index in the current state and the current state amount, obtaining the control signal of a driving motor; s3) according to the control signal, controlling the operation of the driving motor so as to make the balance vehicle tend to be balanced; and using a preset evaluation function to evaluate the change in the state amount of the balance vehicle to obtain the compensation evaluation value; s4) according to the compensation evaluation value, updating the state behavior index under the current state; and s5) observing and updating the current state of the balance vehicle and returning to step 1. According to the invention, the method does not require the prior Knowledge of a model, and has the advantages of self-learning and self-adapting. And the method can be applied flexibly.