Object control method, device and computer-readable storage medium

Abstract

The embodiment of the present application discloses an object control method, device, and computer-readable storage medium. By acquiring the first action network model of the object of each occupation type after training under a single object; The object information determines the corresponding second action network model, and shares the network parameters of the first action network model associated with the occupation type of each team object information to the corresponding second action network model; The first state information and the second action network model, output the corresponding first action information; input the first state information and the corresponding first action information of different team object information together into the second discrimination network model, and generate each group The value expectation of the team object information; the model is updated according to the value expectation, and the second target action network model is obtained; each team object information is controlled based on the second target action network model. In this way, the object control efficiency is improved.

Description

technical field [0001] The present application relates to the field of communication technologies, and in particular, to an object control method, apparatus, and computer-readable storage medium. Background technique [0002] With the rapid development of Internet technology, the processing power of intelligent terminal processors has become stronger and stronger, and many game applications have been derived, such as turn-based game applications. The two sides of the battle adopt the method of attacking in turn. Until all combat objects on one side are killed. [0003] In the prior art, most of the non-player characters (Non-Player Character, NPC) in the turn-based game level mode are still attacked by establishing a decision tree, that is, the decision tree algorithm establishes a tree structure by selecting relevant features, The child nodes under each parent node represent all the move strategies of the node. [0004] During the research and practice of the prior art, ...

AI Technical Summary

Problems solved by technology

[0004] During the research and practice of the prior art, the inventors of the present application found that in the prior art, for the decision tree approach, the NPC's moves are often relatively single, and there is no coordination, which is different from that of real players. There are large differences in the way of moves, and the effect of object control is poor

Images