TiFA: A terrain-informed navigation framework for articulated tracked robots in rescue missions

Abstract

Articulated tracked robots are well-suited for post-disaster rescue missions, but their navigation in large-scale, complex environments presents significant challenges. Current methodologies often face limitations, including high computational demands, inadequate real-time capabilities, and compromised safety during cross-story maneuvers. This paper introduces a novel navigation framework that addresses these challenges through a hierarchical global planner, a manifold-optimization-based local planner, and an on-manifold model predictive controller. The framework leverages various forms of terrain information extracted from the environment in all planning and control stages to facilitate global path generation, local path optimization, and robust whole-body motion control. With coherent interaction among different modules, the framework ensures reliable navigation performance. Extensive ablation simulations and comparative real-world experiments demonstrate that the proposed framework significantly improves the success rate, safety, and efficiency over state-of-the-art methods, while replacing any key method within our framework leads to a noticeable degradation in performance. Specifically, in large-scale scenarios with complex terrain spanning multiple acres, the hierarchical global planner can generate feasible paths within seconds, reducing the time cost by 78.81%. The manifold-optimization-based local planner effectively ensures obstacle avoidance while fully meeting the maneuvering safety requirements in various typical challenging terrains. The holistic controller enabled the robot to stably and reliably track paths on steep $(\ge 30°)$ and rugged terrain.