Frontier Point Allocation in Multi-Robot Collaborative Exploration with a Hybrid Auction Algorithm

This paper proposes a hybrid auction algorithm that combines neural networks, parallel auction algorithm, and conflict resolution strategy to solve the frontier point task allocation problem in the multi-robot cooperative exploration process under weak communication environment. This paper integrates a trained fully connected neural network into each robot. Each robot takes the frontier point position, its own position, battery remaining power, communication neighbor number and other information as the input of the neural network, and the output of the neural network is the robot's bid for each frontier point. The robots are divided into several groups according to the communication topology structure. Each robot group adopts an improved parallel auction algorithm to perform the frontier point task allocation within the group. However, unlike the traditional parallel auction algorithm, each robot's bid is not obtained by the cost function, but by the neural network. And before sending the task allocation result to the robot that executes the task, several robot groups' auctioneers need to ensure the rationality and consistency of the task allocation through the conflict resolution strategy. After the conflict resolution stage is over, each auctioneer broadcasts the task allocation result within their own group. In the experiment, We compared our method with the Greedy algorithm. Our method has more advantages as the scene gets larger and is more robust than the Greedy algorithm.