📌 July 6 & 7 — 8:00 AM (PDT)
📺 Streamed LIVE on YouTube, Twitter, and Facebook.
✨ Check-out all the info and register on our website https://px4.io/virtual-2020/
Back To Schedule
Tuesday, July 7 • 11:30am - 11:45am
Autonomous Exploration Path Planning in High-risk Environments using Aerial Robots

Log in to save this to your schedule, view media, leave feedback and see who's attending!

Feedback form is now closed.

In this talk we present a library of exploration path planning algorithms for aerial robots operating in high-risk settings such as underground environments. Given a previously unknown space, the exploration planners provide collision-free trajectories to navigate the space and build a complete map of the environment autonomously. The set of presented planners emphasize different aspects, such as on large-scale narrow-width subterranean environments, saliency-aware exploration, uncertainty-aware navigation and more. Among the multiple planners in the presented set, we will particularly emphasize on a novel graph-based planning framework purposefully built for aerial robots in subterranean environments such as underground mines, metropolitan infrastructure, and caverns. Given their often large-scale topology and complex network of branches, a bifurcated approach consisting of local and global planning layers is designed. The local planner employs a rapidly-exploring random graph to efficiently identify exploration paths within a local subspace. The global planner is developed as another layer to reposition the robot towards other unexplored areas over the global map when the robot reaches a dead-end situation. This planning algorithm is evaluated extensively in a variety of underground mines and bunkers in the U.S. and in Switzerland. The framework is also being used by the CERBERUS team in DARPA Subterranean Challenge which focuses on innovative robotics solutions for exploration and search missions in underground environments (https://www.subtchallenge.com/). Secondly, we will also focus on two multi-objective planners that co-optimize for exploration (extrinsic goal) and the intrinsic goals of localization uncertainty minimization and salient region re-observation. All the planners are interfaced via the Robot Operating System, a subset is open-sourced already, and to a great extent have been tested with PX4-enabled platforms.

A list of demonstration videos from our experiments and field deployments could be found at https://www.youtube.com/playlist?list=PL3qlfAXmuOFIr-m8ScWJXdDO88_feibTq

Software related to the proposed planners that are already open-sourced is available at
- Graph-based Exploration Planner for Subterranean Environments: https://github.com/unr-arl/gbplanner_ros (in progress)
- Motion-primitives Based Planner for Fast & Agile Exploration: https://github.com/unr-arl/mbplanner_ros (in progress)
- Visual Saliency-aware Exploration Planner: https://github.com/unr-arl/rhem_planner
- Uncertainty-aware Receding Horizon Exploration and Mapping: https://github.com/unr-arl/vseplanner 
- Tutorial: https://www.autonomousrobotslab.com/subtplanning.html

avatar for Tung Dang

Tung Dang

Graduate Researcher, University of Nevada, Reno
I am Tung Dang, PhD candidate at the University of Nevada, Reno. Currently, my research is focused on the informative path planning for autonomous exploration in large-scale subterranean environments.

Tuesday July 7, 2020 11:30am - 11:45am PDT