Mechanical arm control system calibration method based on deep reinforcement learning, storage device and electronic equipment

The invention discloses a mechanical arm control system calibration method based on deep reinforcement learning, a storage device and electronic equipment. The method specifically comprises the steps that system errors are defined, wherein the system errors comprise camera pose errors and mechanical arm coordinate system errors; secondly, calibrating the pose error of the camera, acquiring multiple pieces of pose information and corresponding real images of the target object from a reality system, acquiring simulation images with the same number in a simulation system according to the pose information, processing the two groups of images by using a deep reinforcement learning model, and determining the pose error of the target object according to the processing result; and the camera pose error calibration of the two sets of systems is guided. And finally, the coordinate system error of the mechanical arm is calibrated, the calibration process is similar to calibration of the camera pose error, however, the t