FAR Planner: Fast, Attemptable Route Planner using Dynamic Visibility Update

FAR is a path planning algorithm based on visibility graph. The method is capable of efficiently handling unknown and partially known environments. The method uses a two-layer implementation of the visibility graph, constructing the local layer in the vicinity of the vehicle at each data frame and dynamically merging the local layer into the global layer for updates. In the video, we show the planner’s ability to navigate through unknown environments, fast replan and re-route once reached a “dead-end”. FAR planner has been integrated into Autonomous Exploration Development Environment. The two repositories form a full stack of planning algorithms for ground vehicle navigation to promote research in navigation autonomy. Our work has been submitted to ICRA 2022. Open-sourced Code: https://github.com/MichaelFYang/far_planner Autonomous Exploration Development Environment: https://www.cmu-exploration.com Matterport3D: https://niessner.github.io/Matterport/