ETPP: An Efficient Traversal Path Planning Framework for Autonomous Robot Exploration in Unknown Environments

Autonomous robot exploration in unknown environments is often constrained by the inefficiency of locally optimal strategies that evaluate frontier candidates individually for next-best-view (NBV) selection. To address this limitation, we propose an Efficient Traversal Path Planning (ETPP) framework. During exploration, frontiers are dynamically generated, and a roadmap representing the environmental backbone topology is incrementally constructed. Subsequently, a traversal path planning module is introduced, which employs an adaptive iterative cyclic crossover COOT algorithm to determine the global traversal path across frontiers. This global path is then combined with a local exploration path planned via an information entropy-based algorithm. For NBV selection, we design a distance–orientation evaluation function. Finally, robot navigation within the unknown environment is achieved using an Angle-based Pedestrian Grid Soft Actor–Critic (APG-SAC) motion controller. Mutual information is utilized for online NBV updates to minimize exploration cost. Simulation results demonstrate that ETPP reduces path length and time consumption by 15.95% and 21.04%, respectively, compared to state-of-the-art methods. Real-world experiments further validate the practical feasibility of the proposed framework.