HARDNAV - Simulator for Benchmarking Robust Navigation and Place Recognition in Large, Confusing and Highly Dynamic Environments

Tomáš Musil, Matěj Petrlík, Martin Saska

 

This email address is being protected from spambots. You need JavaScript enabled to view it.   This email address is being protected from spambots. You need JavaScript enabled to view it.
  github.com/MrTomzor/navigation_unity_testbed
  video showcase for IROS2023

 

Abstract — We present a novel simulator called HARDNAV for developing and benchmarking robust autonomous navigation and place recognition. The simulator is designed to effortlessly simulate challenging scenarios and environmental features for single-mission autonomy such as dynamic objects and lights, featureless areas, sensor corruption, and for long-term autonomy such as visibility or structural and topological changes and large-scale unstructured environments. Additionally, we propose replicable benchmarks of active place recognition, and of multi-session navigation, specifically for a kidnapped robot return home mission type, and discuss other challenging benchmarks possible in our simulator. We opensource the code for the simulator and provide scripts and tutorials to easily design multi-session experiments. We hope the simulator will serve the robotics and AI community to develop robust spatial intelligence methods.