A Receding Horizon Approach to Resource Allocation and Data Harvesting
Committee Members: Advisor: Christos Cassandras, SE/ECE; David Castañón, SE/ECE; Calin Belta, SE/ME
Abstract: This thesis focuses on problems of cooperative nature where a group of mobile agents cooperate towards a common objective. We seek alternative solution methods for problems whose most commonsolutionmethodssufferfromcombinatorialandstochasticcomplexity. Webuildonexisting methods for resource maximization missions and focus on two main problems, the resource allocation problem and the data harvesting problem.
The nature of modern cooperative systems requires real-time decision making in a dynamic on-line setting. Using a Cooperative Receding Horizon (CRH) control scheme and adjusting it to fit the problems at hand, we are able to solve these problems in a computationally efficient way which allows real-time applications. The receding horizon approach calculates control values by solving a problem which optimizes a cost function over a limited window of time into the future. This optimization problem is solved repeatedly as the state of the system changes, including random events.
Most real-life problems have some source of stochasticity and require solution methods to be able to cope well with such randomness. Static solution methods do not cope well with stochasticity and the hedge and react approach of the CRH control scheme is a viable alternative for stochastic problems.
To verify the effectiveness of the algorithms developed in this thesis, we built a simulation environment in software as well as a hardware testbed. These simulation environments enable us to build a cooperative system and evaluate its performance for various kinds of scenarios. The performance of the algorithms is compared to known existing methods where that is applicable.