A Decision-Theoretic Planner with Dynamic Compound Reconfiguration for Distributed Real-Time Applications

Book One

Volume

Proceedings of the AAAI Conference on Artificial Intelligence, 21

AAAI Member Abstracts

Middleware is increasingly being used to develop and deploy components in large-scale distributed real-time and embedded (DRE) systems, such as the proposed NASA sensor web composed of networked remote sensing satellites, atmospheric, oceanic, and terrestrial sensors. Such a system must perform sequences of autonomous coordination and heterogeneous data manipulation tasks to meet specified goals. The efficacy and utility of the task sequences are governed by dynamic factors, such as data analysis results, changing goals and priorities, and uncertainties due to changing environmental conditions. These task sequences can be implemented in DRE systems using component middleware, which automates remoting, lifecycle management, system resource management, and deployment and configuration. To support such DRE systems, we developed the Spreading Activation Partial Order Planner (SA-POP) for dynamic (re)planning under uncertainty. SA-POP operates on a spreading activation network, whose structure captures the functional relationships between tasks (implemented as components) and system/environmental conditions (including goals). In this network, utility values capture the importance of desired goals, and probabilities capture the likelihood of tasks succeeding under different conditions. We use a novel partial-order planning and scheduling algorithm to extract a plan for a complete application from this network. We combine SA-POP with the Resource Allocation and Control Engine (RACE), which is a reusable component middleware framework including resource allocation and control algorithms to enforce quality of service (QoS) requirements. The combination of SA-POP and RACE promises an efficient and scalable architecture supporting autonomy in DRE systems operating in dynamic and uncertain domains.