Autonomous Informative Path Planning for Aerial and Legged Robots
Loading...
Authors
Dharmadhikari, Mihir
Issue Date
2021
Type
Thesis
Language
Keywords
Aerial Robots , Legged Robots , Path Planning , Robotics , Subterranean
Alternative Title
Abstract
Subterranean environments, such as cave networks and underground mines, often involve complex, large-scale and multi-branched topologies with complex terrain, and perceptually degraded conditions. Such settings present challenges to autonomous information gathering for aerial and ground platforms alike. In response to the above, this study details a set of path planning methods tailored to the problem of autonomous exploration using aerial and ground robots, including the task of combined volumetric mapping and simultaneous visual search. The first part of this thesis presents an exploration path planner unified across aerial and ground robots. The method builds upon a graph-based exploration path planning scheme and focuses on a) enabling navigation in possibly multi-level environments with uneven terrain and steep slopes, as well as b) ensuring lightweight computational performance. A set of field experiments in various underground environments using legged and flying robots are presented to evaluate the proposed method. Furthermore, the method is open-sourced and a brief description of the software stack and its use is provided. Notably, the presented method contributed into the winning performance of Team CERBERUS at the DARPA Subterranean Challenge Finals.
The second half of the thesis presents a hypergame-based adaptive behavior planner for combined exploration and visual search of unknown environments. As volumetric exploration and visual search are possibly conflicting objectives involving different sensors, a hypergame formulation between the robot and the environment is proposed that allows the robot to select an appropriate behavior based on the local environment geometry, the encountered visual conditions, and a self-evaluation of its performance. A detailed evaluation is presented and includes a) a comparative study against a pure exploration planner both in simulation and in a real experiment inside an indoor environment, alongside b) a field test in an underground mine.