An Ensemble Approach Towards Dynamic Task Assignment for Many Robot Systems

In the last ten years, there has been significant interest in applying a swarming paradigm towards the control and coordination of large teams of robots where individual robots are programmed with simple but identical behaviors that can be realized with limited on-board computational, communication, and sensing resources. While there has been some success in the development of scalable distributed control strategies, most have been limited to the motion coordination and/or consensus forming techniques developed for teams of homogeneous robots. While swarming paradigms lend themselves to sophisticated analysis, these analyses tend to ignore effects when the team is split into smaller subcomponents and assigned to execute different tasks. The allocation of a large team of robots to individual tasks inevitably leads to smaller numbers of locally interacting groups of robots. The presence of these smaller scales motivates the need for multi-robot-style techniques that remain amenable to whole-team analysis.;Towards this end, this thesis presents an ensemble modeling and control framework towards the design of distributed control strategies to address the dynamic allocation of a team of robots to a collection of spatially distributed tasks. The problem of efficient allocation of mobile sensing resources in a dynamic and uncertain environment is relevant for applications like large scale environmental monitoring, surveillance, and distribution warehouse automation. In these applications , the team must have the ability to autonomous allocate themselves and execute the various tasks to ensure task completion and/or provide coverage which may be affected by robot failures or changes in the environment. Different from existing multi-task (MT) robots, single-robots (SR), time-extended assignment (TA) strategies, the techniques described in this work rely on the development of an appropriate macroscopic description of the team dynamics in the development of individual control policies that do not require robots to execute complex bidding schemes. The advantage is a framework that simultaneously address the allocation and controller synthesis problems for the team. The result are decentralized robot control strategies that are invariant to team size and can be implemented with minimal inter-agent communication requires. The proposal begins with a brief literature review and the terminology, notation, and assumptions that provide the foundation of the proposed modeling and control framework. This lays the foundation for the main contribution of this work: the synthesis of distributed control policies for a homogeneous robot team assigned to monitor multiple locations using a macroscopic description of the ensemble dynamics. The resulting controllers are agent-level control policies that can shape the distribution of the robot populations across the various sites that can be implemented with minimal inter-agent wireless communication. This is important since it is often impractical for resource constrained agents to have the ability to transfer complex state information across large distances. The document concludes with a proposed list of specific avenues of investigation and an execution plan to ensure the completion of the thesis.