Multi-Agent Task Assignment

Background:

Multi-agent Task Assignment is a prerequisite for many autonomous, multi-agent systems because of the need to intelligently assign tasks to teammates for maximizing efficiency. The focus of the research community:

  • Developing methods that optimize performance in relatively controlled settings
  • Assuming only a minor amount of uncertainty
  • Expecting a low probability of failure

In many real-world missions, however, uncertainty and failure are paramount concerns because they naturally degrade performance, can have cascading effects on agents and future task allocations and could prevent successful completion of a mission altogether. To overcome these challenges, we seek task allocations that enable resilient operations by considering more expressive models and improve the probability of mission success rather than strict optimization of some classical objective function, e.g., minimizing the makespan.

Our laboratory’s focus:

We are developing a novel heuristic for MRTA that removes common Markovian assumptions and incorporates state and uncertainty modeling in order to improve performance. When tested in a simulated search and rescue mission using real-world autonomous navigation data, our novel heuristic shows a considerable improvement in mission performance by modeling probabilistic task failure, agent failure, failure detection, and non-Markovian states.

Contingency-Aware Task Planning

Background:

Complex logistics support missions require the execution of spatially separated information gathering and situational awareness tasks by a team of robots. Planning strategies for such missions must consider the formation of effective coalitions among available robots and the assignment of tasks to robots with the goal of minimizing the expected mission completion time. The occurrence of unexpected situations that adversely interfere with the execution of the mission may require the execution of contingency tasks so that the originally planned tasks may proceed with minimal disruption.

Our laboratory’s focus:

We are developing a proactive approach to incorporate potential contingency tasks. We initially generate a plan for nominal mission execution. Detection of a new potential contingency task leads to replanning. The planner must update the mission execution plan based on the probability of the potential contingency task to impact the mission tasks. It may either immediately incorporate the contingency task in the mission plan or defer its incorporation until its information changes. We have demonstrated that the proactive approach to contingency task management outperforms both the conservative and reactive approaches.

Task-Agent Assignment for Mobile Manipulators

Background:

Multi-arm mobile manipulators can be represented as a combination of multiple robotic agents from the perspective of task-assignment and motion planning. Depending upon the task, agents might collaborate or work independently. Integrating motion planning with task-agent assignment is a computationally slow process as infeasible assignments can only be detected through expensive motion planning queries.

Our laboratory’s focus:

We have developed three speed-up techniques for addressing this problem:

  1. Spatial constraint checking using conservative surrogates for motion planners
  2. Instantiating symbolic conditions for pruning infeasible assignments
  3. Efficiently caching and reusing previously generated motion plans.

We have shown that the developed method is useful for real-world operations that require complex interaction and coordination among high-DOF robotic agents.