Generating Informative Trajectories for Robots Persistently Monitoring Unknown Environments

From DRLWiki

Revision as of 18:10, 27 November 2012 by Soltero (Talk | contribs)
(diff) ← Older revision | Latest revision (diff) | Newer revision → (diff)
Jump to: navigation, search

Approach

Building upon previous work from Mac Schwager, et. al, we developed a controller to manipulate a collection of waypoints defining a robot's path into an informative path. By informative path, we mean a path that allows the robot to sense locations in the environment that contain some kind of information; we refer to these locations as points of interest. This controller introduces to primitives to the movements of the robot's waypoints: 1) go towards its Voronoi centroid, and 2) stay close to the next and previous waypoint in the robot's path. These two primitives make the waypoints move in a way that generates a closed path that goes through the points of interest in the environment and avoids going through locations that do not contain any information.

In order to treat unknown environments, an adaptation law was generated, which allows the robot to estimate the environment description by making point measurements and integrating these measurements. The controller then uses the estimated environment description to determine how the waypoints should move.

We apply our informative paths to persistent sensing by combining our paths with a speed controller from previous work, which calculates the speed along each point in the path in order to perform a stable persistent task. A persistent task consists of a robot maintaining an accumulation function in the environment bounded everywhere and at all times. This accumulation function grows at every point in the environment at a linear rate and is consumed by the robot only when the robot's sensor is covering the point in the environment. By combining our informative paths with the speed controller, we generate informative trajectories that allow the robot to perform a persistent sensing task.

We extended our controller to the multi-robot case where the robots have to work together to perform the persistent sensing task. This allows us to execute these persistent tasks in environment where a single robot is unable to do so.


Simulated initial configuration for a single-robot case. Blue lines represent Voronoi partitions, red dots represent points of interest, and the black line represents the path.


Simulated final configuration for a single-robot case. The final path only goes through the points of interest, providing good sensing locations for the robot.
Initial configuration for a 2-robot experiment. Green dots represent points of interest and the size of the green dots represents the accumulation function that needs to be consumed, the red and blue lines represent the paths for each robot.
Configuration after 50 iterations in the 2-robot experiment
Final configuration for the 2-robot experiment. The final path only goes through the points of interest, providing good sensing locations for the robot. Combined with the speed controller, these robots perform a stable persistent sensing task

Publications

Single-robot case:

D. E. Soltero, M. Swchager, and D. Rus, “Generating informative paths for persistent sensing in unknown environments,” in Intelligent Robots and Systems (IROS), 2012 IEEE/RSJ International Conference on, 2012.



Multi-robot case:

Submitted to ICRA 2013

People

Daniel E. Soltero

Mac Schwager

Daniela Rus

Personal tools