Communication-Constrained Multi-Robot Coordination

Work began in 2019.

We propose to create and implement methods that enable a team of robots to collaboratively carry-out a task, such as searching an area for an moving evader. Whereas previous methods assume unlimited, reliable, and instantaneous communication, we assume that communication is both limited and unreliable. At the conclusion of this project, we will demonstrate these capabilities with multiple MAGIC robots pursuing an evader within a building — essentially playing “tag” — but where communication between the robots is always limited (bits per second) and often denied entirely.

Previous methods have assumed unlimited communication available between the robots, but we propose to explicitly plan around limited/denied communication. This work will build upon our previous work in Multi Policy Decision Making (MPDM). MPDM allows robots to choose from multiple policies even in complex environments. We will apply MPDM to the problem of deciding how to search the environment for the evader. At its core, MPDM allows a system to dynamically select between several policies. Each policy can be viewed as a control policy which maps the current perceptual input to a robot action (e.g., rotate left). The core idea behind MPDM is to use a behavioral level simulator, running on the robot in real time, to “play out” the current situation using each of the candidate policies.

Prior Related Publications:

• Dhanvin Mehta, Gonzalo Ferrer and Edwin Olson. Backprop-MPDM: Faster risk-aware policy evaluation through efficient gradient optimization. Proceedings of the IEEE International Conference on Robotics and Automation (ICRA), May 2018.
• Edwin Olson, Johannes Strom, Robert Goeddel, Ryan Morton, Pradeep Ranganathan and Andrew Richardson. Exploration and Mapping with Autonomous Robot Teams. Communications of the ACM, March 2013.
• Ryan Marcotte and Edwin Olson. Adaptive forward error correction with adjustable-latency QoS for robotic networks. Proceedings of the IEEE International Conference on Robotics and Automation (ICRA), May 2016.