Learning from animals: How to Navigate Complex Terrains

• DOI: 10.1371/journal.pcbi.1007452
• Publication Date: 2020-01-09T13:25:27Z
• Authors: Henghui Zhu; Hao Liu; Armin Ataei; Yonatan Munk; Thomas Daniel; Ioannis Ch. Pasch...
• Keywords: Robustness; Optical Flow

We develop a method to learn bio-inspired motion control policy using data collected from hawkmoths navigating in virtual forest. A Markov Decision Process (MDP) framework is introduced model the dynamics of moths and sparse logistic regression used parameters data. The results show that do not favor detailed obstacle location information navigation, but rely heavily on optical flow. Using learned moth as starting point, we propose an actor-critic learning algorithm refine obtain can be by autonomous aerial vehicle operating cluttered environment. Compared with moths' policy, integrates both compare performance these two policies terms their ability navigate artificial forest areas. While optimized adjust its outperform moth's each different terrain, exhibits high level robustness across terrains.