FLTPlanner: Fast Legged Trajectory Planner for Real-Time Foothold Reachability Analysis and Swing-Trajectory Generation

Legged robots today face significant challenges when navigating complex environments, requiring precise real-time decisions for foothold selection and foot trajectory planning. While various planners have been developed for swing trajectory planning, many are either too time-consuming or overly simplified for handling intricate environments. This paper introduces FLTPlanner, a novel approach for real-time foothold reachability analysis and trajectory generation. For each swing-trajectory planning task, we first assess foothold reachability using a predefined feasible domain and grid map. We then initialize a heuristic trajectory that is close to feasible solutions. This initial trajectory undergoes MINCO trajectory optimization, incorporating collision and limit constraints, to achieve the optimal path while avoiding local minima. We validated FLTPlanner on the Elspider4 Air robot, both in simulations and real-world tests. The robot consistently outperformed other planners in complex terrains and confined spaces.

IEEE Robotics and Automation Letters(Under Review)