Traversability-aware path planning in dynamic environments

Planning in environments with moving obstacles remains a significant challenge in robotics. While many works focus on navigation and path planning in obstacle-dense spaces, traversing such congested regions is often avoidable by selecting alternative routes. This paper presents Traversability-aware FMM ( Tr-FMM), a path planning method that computes paths in dynamic environments, avoiding crowded regions. The method operates in two steps: first, it dis-cretizes the environment, identifying regions and their distribution; second, it evaluates the traversability of regions, aiming to minimize both obstacle risks and goal deviation. The path is then computed by propagating the wavefront through regions with higher traversability. Simulated and real-world experiments demonstrate that the approach ensures significant safety by keeping the robot away from obstacles while minimizing excessive goal deviations. Introduction Robots operating without direct human supervision or intervention in everyday life are becoming increasingly common. Consequently, moving in spaces shared with humans emerged as a significant challenge in robotics [1]. Typical examples of such environments include indoor settings like stores, warehouses, and airports [2]. In these crowded or busy environments, people often move unpredictably or without paying sufficient attention to robots, potentially leading to collisions or deadlock situations from which a robot cannot recover [3]. Therefore, it is crucial that robots are capable of avoiding such situations, where people are seen as dynamic obstacles needed to be avoided. Classic and widely used navigation techniques, such as DW A [4] and elastic bands [5], struggle in the aforementioned situations. DW A is designed for static scenarios, while elastic bands are not suited for highly dynamic and crowded environments. Navigation methods that account for dynamic obstacles, such as VO [6], RVO [7], and ORCA [8], are designed as local planners for maneuvering among people or moving obstacles, rather than as global planners for such scenarios. More recent approaches, often based on advanced learning techniques [9][10][11], demonstrate higher success rates in avoiding collisions. All these techniques are most useful when the robot is already inside a crowd or has no choice but to pass through one, accepting the potential risk of collision.