Prev-Talk | Next-Talk | All-Talks | Talks-Sorted | MOOS-DAWG'19

Talk-22: Obstacle Avoidance on the Remote Explorer 4

Paul Robinette, Michael Novitzky, Austin Wang, Michael DeFilippo, Michael Sacarny, Michael R. Benjamin

Autonomous surface vehicles must be able to avoid both stationary obstacles and moving vessels as they navigate waterways to accomplish their mission goals. On the open ocean, this is a relatively simple procedure, but confined waterways such as rivers and harbors present special challenges. For the last several years, we have been developing the Remote Explorer (REx) 4 into an autonomous vehicle capable of avoiding various hazards such as buoys, bridge pylons, kayaks, sailboats, and powerboats on the Charles River and in Boston Harbor. This talk will discuss our current progress on this effort as well as our processing pipeline that enables obstacle avoidance.

In Fall 2018, we demonstrated REx 4 avoiding a stationary obstacle (in this case, the powerboat R/V Philos maintaining a position) using our Velodyne HDL-32 LIDAR. LIDAR returns were ingested in a MOOS app and provided to the MOOSDB as binary data. This binary data was then fed into a point cloud using Point Cloud Library (PCL). The point cloud was filtered for blobs of points that could potentially indicate obstacles. The nearest obstacle was selected and its centroid was sent to pObstacleManager so that the vessel could navigate around it. This was successfully demonstrated by commanding REx 4 to go to a waypoint behind R/V Philos such that REx 4 needed to navigate around the powerboat.

This year, we use a ROS-based perception system that provides a set of obstacles and their key points to pObstacleManager through an interface described in another talk. This allows us to leverage existing ROS nodes for LIDAR processing while also using the robust obstacle avoidance behaviors in MOOS-IvP. We intend to demonstrate arbitrary obstacle avoidance such that REx 4 can avoid bridge pylons and buoys before the end of the summer.

Categories:

  • MOOS-IvP
  • Cyber Security