<p>In multi-robot 2D path planning, the persistent challenges of insufficient conflict detection accuracy under constrained spaces and non-unit velocity conditions, along with conflict resolution algorithms' tendency to fall into local optima, remain formidable obstacles. To address these issues, this study proposes a dynamic threat assessment model-based conflict resolution method for multi-robot systems. By employing two-dimensional cellular automata for spatial discretization modeling, we transform the conflict resolution problem into a state transition challenge in discrete space–time domains. Building upon this framework, we design local rules integrating both positional and velocity information, establishing a collision classification-based conflict detection model that enables precise real-time identification and categorical processing of potential conflicts. Furthermore, through dynamic threat minimization combined with dual-iteration state analysis of surrounding cellular automata, we develop evaluation criteria for lane-changing initiation and termination conditions, ensuring effective conflict resolution in confined spaces while maintaining minimal path cost. Extensive simulation experiments conducted in textile workshop-like environments demonstrate the proposed method's marked advantages over the conflict-based search (CBS) algorithm in conflict detection metrics, path computation time, and trajectory length. Notably, the success rate consistently outperforms CBS by 20–50% with increasing robot numbers. These findings validate the method's efficiency and robustness in multi-robot path planning. Our research provides novel insights and strategies for spatiotemporal conflict resolution in multi-robot systems with non-unit velocities. Future investigations will focus on resolving complex conflict patterns and incorporating dynamic obstacle scenarios.</p>