Decision Making in Large-Scale Domains: A Case Study

Planning under uncertainty attracted significant attention in AI and in other fields. To overcome computational problems associated with Markov Decision Processes (MDPs) in large scale domains researchers often take advantage of structural properties and look for approximately optimal solutions. DTGolog, a decision-theoretic agent programming language based on the situation calculus, was proposed to ease some of the computational difficulties by using natural ordering constraints on execution of actions. Using DTGolog, domain specific constraints on the set of available policies can be expressed in a high-level program and this program helps to reduce significantly computation required to find a policy optimal in this set. Our paper explores whether the DTGolog framework can be used to evaluate different designs of a decision making agent in a large real-world domain. Each design is understood as combination of a template (expressed as a Golog program) for available policies and a reward function. To evaluate and compare alternative designs we estimate the probability of goal satisfaction for each design. As a domain, we choose the London Ambulance Service (LAS) case study that is well known in software engineering, but remains unknown in AI. In our paper we demonstrate that DTGolog can be applied successfully to quantitative evaluation of alternative designs in terms of their ability to satisfy a system goal with a high probability. We provide a detailed axiomatization of the domain in the temporal situation calculus with stochastic actions. The main advantage of this representation is that neither actions, not states require explicit enumeration. We do an experimental analysis using an on-line implementation of DTGolog coupled with a simulator that models real time actions of many external agents.

Authors

• Mikhail Soutchanski
• Huy Pham
• John Mylopoulos

Keywords

No keywords are indexed for this paper.