Enhanced CBBA with Multi-Condition Evaluation for Multi-Robot Collaborative Exploration

In multi-robot collaborative exploration of unstructured environments, redundant exploration often occurs, leading to significant consumption of computational resources and thus reducing exploration efficiency. To address this issue, this paper proposed the EC-MCE algorithm, an enhancement of the Consensus-Based Bundle Algorithm (CBBA) algorithm, which integrates a fully connected neural network to evaluate five key factors: travel cost, information gain, loop closure likelihood, revisit probability, and deviation for local benefit evaluation. This approach ensures the efficient allocation of boundary points, which in turn ensures the efficient completion of exploration tasks. The effectiveness of this method was verified on the Airsim simulation platform, with experimental results showing that compared to traditional methods, this approach reduced the total exploration time by about 10% and decreased the number of repeated area explorations by about 20%. These improvements significantly enhance the exploration efficiency of the system while ensuring the safe navigation of multiple mobile robots in unstructured environments.