The invention provides a task
allocation method based on an improved DQN
algorithm, and belongs to the field of task allocation, and the method comprises the steps: creating a DQN
algorithm model, creating a training model, generating scene parameters of a current scene, defining an agent classification standard, carrying out the
pairing of agents and actions through combining a task allocation rule, and calculating a local reward and a global reward. The method comprises the steps of obtaining a
reward value used for DQN network updating, updating a node value of an evaluation network based on the
reward value used for DQN network updating to obtain an updated evaluation network, updating parameters of a target network based on the updated evaluation network to enable the target network to approach the evaluation network, and obtaining a model after current scene optimization. According to the method, different rules and decision strategies are automatically selected in a dynamic environment based on an improved deep
Q learning algorithm of the rules, so that the
system can continuously adjust decisions according to real-time scene information, and thus a more efficient task allocation decision can be made in the face of changeable actual scenes.