Reinforcement learning constraint control method based on single-arm manipulator

The invention provides a reinforcement learning constraint control method based on a single-arm manipulator, and the reinforcement learning constraint control method adopts a general transformation function and error transformation to eliminate feasibility conditions caused by related constraint problems in order to meet asymmetric full-state time-varying constraint requirements. According to the scheme, a nonlinear filter related to self-adaptive parameters is constructed to overcome the problem of calculation explosion existing in a backstepping method, and the asymmetrical full-state time-varying constraint characteristic is brought into an optimal framework to guarantee the optimality of a single-arm manipulator model. According to the scheme, the problem that the control performance is poor due to the fact that an existing single-arm manipulator model is affected by the external environment and the like due to the complexity of the working environment is solved, and the purposes of full-state time-varyin