Brandon J. Moore

Distributed Task Assignment for Mobile Agents

This note demonstrates how the distributed auction algorithm can be modified to assign mobile agents to spatially distributed tasks despite communication delays and the fact that agent movement may cause the benefit associated with each possible agent-task assignment to vary during the execution of the algorithm. Bounds on the convergence time of the algorithm and the sub-optimality of the resulting solution are provided. Monte Carlo simulations are provided to show the conditions under which the modified distributed auction can outperform centralized calculation

Distributed Balancing of AAVs for Uniform Surveillance Coverage

This paper addresses the problem of enabling a group of autonomous air vehicles to provide surveillance coverage for an area significantly larger than their communication radius. Our formulation spatially decomposes the overall surveillance mission into subtasks and we develop a distributed cooperative control algorithm that continuously reassigns AAVs to these subtasks based on only local information in order to achieve the most balanced distribution of AAVs possible within a finite time interval of known bound. Various applications are discussed and simulations are included to illustrate convergence dynamics as well as to measure practical performance as a function of problem parameters.

Brief paper: Optimal allocation of heterogeneous resources in cooperative control scenarios

This paper introduces a mathematical framework for the study of resource allocation problems involving the deployment of heterogeneous agents to different teams. In this context, the term heterogeneous is used to describe classes of agents that differ in the basic functions they can perform (e.g., one type of agent searches for targets while another type engages those targets). The problem is addressed in terms of optimization using concepts from economic theory and the proposed algorithm was designed to ensure asymptotic stability of the optimal solution.

Distributed Coordination Strategies for Wide-Area Patrol

This paper addresses the problem of enabling a group of autonomous vehicles to effectively patrol an environment significantly larger than their communication and sensing radii. The environment is divided into smaller areas and special coordinator vehicles are designated to control the transfer of the other vehicles from one area to another. Our past work showed that by organizing the areas and coordinators into a ring topology, we could design a control algorithm that globally balanced the number of vehicles in all areas within a bounded length of time. This paper extends those results to a much broader class of area-coordinator topologies and this added flexibility can be used in implementation to reduce the time it takes to attain the globally balanced state.

Spatial balancing of autonomous vehicle resources

This paper addresses the problem of achieving a balanced distribution of autonomous vehicles across a locally connected topology of areas. Motivated by a cooperative autonomous air vehicle mission scenario that requires the performance of extended surveillance over distances large enough to impede vehicle-to-vehicle communication, the proposed algorithm allows the group to achieve the desired distribution of vehicles through the use of a special set of coordination vehicles

Stable emergent agent distributions under sensing and travel delays

In order for a team of cooperating agents to achieve a group goal (such as searching for targets, monitoring an environment, etc.) those agents must be able to share information and achieve some level of coordination. Since realistic methods of communication between agents have limited range and speed, the agents¿ decision-making strategies must operate with incomplete and outdated information. Moreover, in many situations the agents must travel to particular locations in order to perform various tasks, and so there will also be a delay between any particular decision and its effect. In this paper we develop an asynchronous framework that models the behavior of a group of agents that is spatially distributed across a predefined area of interest. We derive general conditions under which the group is guaranteed to converge to a specific distribution within the environment without any form of central control and despite unknown but bounded delays in sensing and travel. The achieved distribution is optimal in the sense that the proportion of agents allocated over each area matches the relative importance of that area. Finally, based on the derived conditions, we design a cooperative control scheme for a multi-agent surveillance problem. Via Monte Carlo simulations we show how sensing and travel delays and the degree of cooperation between agents affect the rate at which they achieve the desired coverage of the region under surveillance.