Introduction

Advances in robot autonomy have led to a decrease in the necessity of strict human supervision of robots. With the popularity of robotics technology on the rise, we are now able to develop human-robot collaborative systems where the task is primarily executed by a number of semi-autonomous robots requiring intermittent assistance from a human teammate. This can be seen in a variety of applications such as warehouse operations, search-and-rescue, and in social settings.

Challenges in Human-Robot Collaboration

However, identifying which robot to assist in an uncertain environment is a challenging task for human operators. As the number of robots increases, it becomes increasingly difficult for operators to maintain awareness of every robot, which can cripple the system's performance. Therefore, human operators can benefit from having a decision support system (DSS) that advises which robots require attention and when.

Proposed Solution: Decision Support System

In this project, we present such a DSS for a multi-robot system comprising a fleet of semi-autonomous robots with multiple human operators available for assistance if and when required. The system operates in a city block-like environment, with robots moving from start to goal locations. While navigating, a robot passes through a series of waypoints, each characterized by a different probability of success in progressing to the next waypoint. There is also a possibility that the robot may fail at a task and get stuck in a fault state from which assistance from a human operator is required to continue.

The DSS is designed to efficiently allocate operators to the robots by decomposing the complete multi-robot problem into individual single-robot problems and computing the Whittle index heuristic. Given the current state of the system, this heuristic can be used to efficiently compute the operator allocation.

Modeling the Problem as a MDP

In addition to the DSS, the project also presents a model of the multi-robot system that is formulated as a Markov Decision Process (MDP). This approach allows for the modeling of the problem as a whole, taking into account the interactions between the different robots and operators, and the probability of success and failure for each robot. The MDP model is an important tool for understanding the problem and for developing new solutions.

Solving the MDP Model

However, solving the MDP model can be computationally challenging, due to the curse of dimensionality. To overcome this challenge, the project proposes the use of approximation methods, such as the Whittle index heuristic, to reduce the computational complexity and make the solution more tractable.

Conclusion

Overall, this project highlights the potential for human-robot collaborative systems and the importance of decision support systems in ensuring efficient and effective operation. As the field of robotics continues to advance, we can expect to see more developments in this area and a greater reliance on human-robot collaboration.

Future work

This project presents a promising solution for addressing the challenges of human-robot collaboration in uncertain environments. However, there is still room for improvement and further research. For example, the approach can be tested and evaluated in more realistic scenarios, such as a warehouse or search-and-rescue operation. Additionally, the DSS could be extended to consider more complex decision making processes, such as incorporating human preferences and priorities. Overall, the goal is to continue to improve the effectiveness and efficiency of human-robot collaboration in various settings.