A strategy and evaluation method for ground global path planning based on aerial images

摘要: Abstract Autonomous robot navigation consists of establishing a sequence actions that guarantee safe and efficient locomotion for robotic system. This work focuses on outdoor ground guided by aerial images. We propose pipeline image processing, traversability estimation, terrain representation route planning capable assessing whether desired target location is reachable or not. also dataset an evaluation method to quantify the average quality generated paths measure how well strategy finds possible routes rejects impossible ones. was applied identify parameter combination maximized path feasibility recognition quality. Our results show proposed identifies correct action in 85.8% test cases takes 2.3 s mapping 100 ms over 300 × 300 m2 areas images size 1000 × 1000 pixels. successfully employed our small scale experiment guide Hydra v3 mini rover.