State estimation and decision control method for autonomous grabbing of non-cooperative target

The invention discloses a state estimation and decision control method for autonomous grabbing of a non-cooperative target, and belongs to the field of autonomous grabbing. Comprising the steps that real-time image data are collected, and the center position of a non-cooperative target is detected in real time through a YOLOv5 algorithm; a PoseCNN algorithm is adopted to estimate the attitude in real time, and a non-cooperative target state real-time estimation strategy is formed; an autonomous grabbing system composed of a mechanical arm end effector (a two-finger clamping jaw) and a target is established as a Markov decision model; adopting a human expert demonstration mode to realize network initialization training and collect supervision data, and storing a state-action data pair; constructing a target function, a network gradient and a loss function, and designing a shared feature extraction unit and a network feature regression part; and training the Actor network by using the supervision data and carry