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

This paper presents a framework for navigating in obstacle-dense environments as posed in the 2016 International Conference on Intelligent Robots and Systems (IROS) Autonomous Drone Racing Challenge. Our framework is based on direct visual servoing and leg-by-leg planning to navigate in a complex environment filled with many similar frame-shaped obstacles to fly through.Our indoor navigation method relies on the velocity measurement by an optical flow sensor since the position measurements from GPS or external cameras are not available. For precision navigation through a sequence of obstacles, a center point-matching method is used with the depth information from the onboard stereo camera. The guidance points is directly generated in threedimensional space using the two-dimensional image data to avoid accumulating the error from the sensor drift. The proposed framework is implemented on a quadrotor-based aerial vehicle, which carries an onboard vision-processing computer for self-contained operation. Using the proposed method, our drone was able to finished in first place in the world-premier IROS Autonomous Drone Racing Challenge.

No abstract

No abstract

No abstract