Trust-based Symbolic Motion and Task Planning for Multi-robot Bounding Overwatch

This project begins 2020.

Bounding overwatch is a military tactic of alternating the movement of coordinated teams to move forward under potential enemy fire. As members in a team take an overwatch posture, other members advance to cover. In robotic bounding overwatch, teams of (semi)autonomous ground vehicles are coordinated to perform such tasks autonomously while a human operator supervises the task and intervenes if necessary. It is therefore important to determine which point to choose for overwatch at each step and whether a robot is trustful enough to perform the overwatch task. Here, the intent is to simultaneously cover the bounding robots and overwatch robots themselves. However, unless each overwatch waypoint along the route is provided prior to the mission and no changes are needed, traditional motion planning and role allocation algorithms do not sufficiently encode the mission intent or trust in the robots to conduct the overwatch tasks. Furthermore, a successful approach will also need to consider the temporal constraints in determining overwatch points beyond just identifying what is reachable.

The goal of the project is to create a trust-based symbolic motion and task planning framework for heterogeneous multi-robot systems (MRS) to perform the bounding overwatch maneuver under temporal logic constraints. Experiment validation of the proposed framework will be performed utilizing the autonomous ground mobile robots in the PI’s lab. Python and Java codes in the Robotic Operating System (ROS) will be generated as the deliverables.

The proposed work will provide a new distributed framework for heterogeneous multi-robot systems (MRS) to collaboratively achieve complex tasks with application to bounding overwatch. Compared to extant approaches, our framework will improve computation efficiency, flexibility of robot assignment, and concurrency of task execution while encoding the intent of mission and robot trust into task and motion planning. It can accommodate robot and environment uncertainties, and is suitable for complex missions in dynamic environments with a large number of heterogeneous robots.

Publications:

• H. Zheng, J. M. Smereka, D. Mikulski, and Y. Wang, “Bayesian Optimization based Trustworthiness Model for Multi-robot Bounding Overwatch”, In Proceedings of the Ground Vehicle Systems Engineering and Technology Symposium (GVSETS), NDIA, Novi, MI, Aug. 10-12, 2021.
• H. Zheng, J. M. Smereka, D. Mikulski, S. Roth, Y. Wang, “Trust-based Symbolic Motion Planning for Multi-robot Bounding Overwatch”, In Proceedings of the Ground Vehicle Systems Engineering and Technology Symposium (GVSETS), NDIA, Novi, MI, November 3-5, 2020.
• H. Zheng, J. M. Smereka, D. Mikulski, and Y. Wang, “Bayesian Optimization based Trust Model for Human Multi-Robot Collaborative Motion Tasks in Offroad Environments”, International Journal of Social Robotics. (under review)
• H. Zheng, J. M. Smereka, D. Mikulski, and Y. Wang, “Trust-based Active Reinforcement Learning for Human Multi-Robot Collaborative Offroad Motion Tasks under Temporal Logic Specifications”, IEEE Transactions on Robotics. (in preparation)

#1.33