Multi-robot task searching method applied to rescue environment

The invention discloses a multi-robot task searching method applied to a rescue environment, which can realize multi-robot cooperation under the conditions that a task position is unknown and sensor noise is uncertain, and complete and efficient task searching is completed. The method comprises the following steps: discretizing a rescue environment, and constructing an object-oriented multi-machine part observable Markov decision model to describe a task search problem according to the rescue environment and robot information participating in search; and each robot judges whether the robot is trapped or not according to the current state of the robot, if the robot is trapped, a guide decision of belief clustering and path planning based on a self-adaptive threshold value is adopted, and otherwise, a decision is made according to the maximum reward of the spanning tree. And each robot executes the current execution action, obtains environment observation and reward values, updates the belief of each object, de