NeBula: Quest for Robotic Autonomy in Challenging Environments; TEAM\n CoSTAR at the DARPA Subterranean Challenge

摘要

This paper presents and discusses algorithms, hardware, and software\narchitecture developed by the TEAM CoSTAR (Collaborative SubTerranean\nAutonomous Robots), competing in the DARPA Subterranean Challenge.\nSpecifically, it presents the techniques utilized within the Tunnel (2019) and\nUrban (2020) competitions, where CoSTAR achieved 2nd and 1st place,\nrespectively. We also discuss CoSTAR's demonstrations in Martian-analog surface\nand subsurface (lava tubes) exploration. The paper introduces our autonomy\nsolution, referred to as NeBula (Networked Belief-aware Perceptual Autonomy).\nNeBula is an uncertainty-aware framework that aims at enabling resilient and\nmodular autonomy solutions by performing reasoning and decision making in the\nbelief space (space of probability distributions over the robot and world\nstates). We discuss various components of the NeBula framework, including: (i)\ngeometric and semantic environment mapping; (ii) a multi-modal positioning\nsystem; (iii) traversability analysis and local planning; (iv) global motion\nplanning and exploration behavior; (i) risk-aware mission planning; (vi)\nnetworking and decentralized reasoning; and (vii) learning-enabled adaptation.\nWe discuss the performance of NeBula on several robot types (e.g. wheeled,\nlegged, flying), in various environments. We discuss the specific results and\nlessons learned from fielding this solution in the challenging courses of the\nDARPA Subterranean Challenge competition.\n