Charles Gretton

A Stochastic Local Search Approach to Vertex Cover

We introduce a novel stochastic local search algorithm for the vertex cover problem. Compared to current exhaustive search techniques, our algorithm achieves excellent performance on a suite of problems drawn from the field of biology. We also evaluate our performance on the commonly used DIMACS benchmarks for the related clique problem, finding that our approach is competitive with the current best stochastic local search algorithm for finding cliques. On three very large problem instances, our algorithm establishes new records in solution quality.

Exploiting probabilistic knowledge under uncertain sensing for efficient robot behaviour

Robots must perform tasks efficiently and reliably while acting under uncertainty. One way to achieve efficiency is to give the robot common-sense knowledge about the structure of the world. Reliable robot behaviour can be achieved by modelling the uncertainty in the world probabilistically. We present a robot system that combines these two approaches and demonstrate the improvements in efficiency and reliability that result. Our first contribution is a probabilistic relational model integrating common-sense knowledge about the world in general, with observations of a particular environment. Our second contribution is a continual planning system which is able to plan in the large problems posed by that model, by automatically switching between decision-theoretic and classical procedures. We evaluate our system on object search tasks in two different real-world indoor environments. By reasoning about the trade-offs between possible courses of action with different informational effects, and exploiting the cues and general structures of those environments, our robot is able to consistently demonstrate efficient and reliable goal-directed behaviour.

Decision-theoretic planning with non-Markovian rewards

A decision process in which rewards depend on history rather than merely on the current state is called a decision process with non-Markovian rewards (NMRDP). In decision-theoretic planning, where many desirable behaviours are more naturally expressed as properties of execution sequences rather than as properties of states, NMRDPs form a more natural model than the commonly adopted fully Markovian decision process (MDP) model. While the more tractable solution methods developed for MDPs do not directly apply in the presence of non-Markovian rewards, a number of solution methods for NMRDPs have been proposed in the literature. These all exploit a compact specification of the non-Markovian reward function in temporal logic, to automatically translate the NMRDP into an equivalent MDP which is solved using efficient MDP solution methods. This paper presents NMRDPP(Non-Markovian Reward Decision Process Planner), a software platform for the development and experimentation of methods for decision-theoretic planning with non-Markovian rewards. The current version of NMRDPP implements, under a single interface, a family of methods based on existing as well as new approaches which we describe in detail. These include dynamic programming, heuristic search, and structured methods. Using NMRDPP, we compare the methods and identify certain problem features that affect their performance. NMRDPPs treatment of non-Markovian rewards is inspired by the treatment of domain-specific search control knowledge in the TLPlan planner, which it incorporates as a special case. In the First International Probabilistic Planning Competition, NMRDPP was able to compete and perform well in both the domain-independent and hand-coded tracks, using search control knowledge in the latter.

Partial weighted MaxSAT for optimal planning

We consider the problem of computing optimal plans for propositional planning problems with action costs. In the spirit of leveraging advances in general-purpose automated reasoning for that setting, we develop an approach that operates by solving a sequence of partial weighted MaxSAT problems, each of which corresponds to a step-bounded variant of the problem at hand. Our approach is the first SAT-based system in which a proof of cost-optimality is obtained using a MaxSAT procedure. It is also the first system of this kind to incorporate an admissible planning heuristic. We perform a detailed empirical evaluation of our work using benchmarks from a number of International Planning Competitions.

Advances in local search for satisfiability

In this paper we describe a stochastic local search (SLS) procedure for finding satisfying models of satisfiable propositional formulae. This new algorithm, gNovelty+, draws on the features of two other WalkSAT family algorithms: R+AdaptNovelty+ and G2WSAT, while also successfully employing a dynamic local search (DLS) clause weighting heuristic to further improve performance. gNovelty+ was a Gold Medal winner in the random category of the 2007 SAT competition. In this paper we present a detailed description of the algorithm and extend the SAT competition results via an empirical study of the effects of problem structure and parameter tuning on the performance of gNovelty+. The study also compares gNovelty+ with two of the most representative WalkSAT-based solvers: G2WSAT, AdaptNovelty+, and two of the most representative DLS solvers: RSAPS and PAWS. Our new results augment the SAT competition results and show that gNovelty+ is also highly competitive in the domain of solving structured satisfiability problems in comparison with other SLS techniques.

Induction of topological environment maps from sequences of visited places

In this paper we address the problem of topologically mapping environments which contain inherent perceptual aliasing caused by repeated environment structures. We propose an approach that does not use motion or odometric information but only a sequence of deterministic measurements observed by traversing an environment. Our algorithm implements a stochastic local search to build a small map which is consistent with local adjacency information extracted from a sequence of observations. Moreover, local adjacency information is incorporated to disambiguate places which are physically different but appear identical to the robots senses. Experiments show that the proposed method is capable of mapping environments with a high degree of perceptual aliasing, and that it infers a small map quickly.

A study of proxies for Shapley allocations of transport costs

We propose and evaluate a number of solutions to the problem of calculating the cost to serve each location in a single-vehicle transport setting. Such cost to serve analysis has application both strategically and operationally in transportation. The problem is formally given by the traveling salesperson game (TSG), a cooperative total utility game in which agents correspond to locations in a travelling salesperson problem (TSP). The cost to serve a location is an allocated portion of the cost of an optimal tour. The Shapley value is one of the most important normative division schemes in cooperative games, giving a principled and fair allocation both for the TSG and more generally. We consider a number of direct and sampling-based procedures for calculating the Shapley value, and present the first proof that approximating the Shapley value of the TSG within a constant factor is NP-hard. Treating the Shapley value as an ideal baseline allocation, we then develop six proxies for that value which are relatively easy to compute. We perform an experimental evaluation using Synthetic Euclidean games as well as games derived from real-world tours calculated for fast-moving consumer goods scenarios. Our experiments show that several computationally tractable allocation techniques correspond to good proxies for the Shapley value.

Verified over-approximation of the diameter of propositionally factored transition systems

To guarantee the completeness of bounded model checking (BMC) we require a completeness threshold. The diameter of the Kripke model of the transition system is a valid completeness threshold for BMC of safety properties. The recurrence diameter gives us an upper bound on the diameter for use in practice. Transition systems are usually described using (propositionally) factored representations. Bounds for such lifted representations are calculated in a compositional way, by first identifying and bounding atomic subsystems, and then composing those results according to subsystem dependencies to arrive at a bound for the concrete system. Compositional approaches are invalid when using the diameter to bound atomic subsystems, and valid when using the recurrence diameter. We provide a novel overapproximation of the diameter, called the sublist diameter, that is tighter than the recurrence diameter. We prove that compositional approaches are valid using it to bound atomic subsystems. Those proofs are mechanised in HOL4. We also describe a novel verified compositional bounding technique which provides tighter overall bounds compared to existing bottom-up approaches.

Decision Support System for a Real-Time Field Service Engineer Scheduling Problem with Emergencies and Collaborations

Repeated replanning with a heuristic for solving a type of vehicle routing problem was used in a dynamic routing and scheduling problem. This problem occurs when field service engineers are assigned a sequence of jobs to attend. The jobs are geographically distributed, and not all jobs to be undertaken are known in advance of planning. This dynamic occurrence of job requests is stochastic. Jobs are assigned an emergency level, which is highest for repair jobs involving a person in danger. In addition, some jobs require two engineers; such jobs are referred to as collaborative. The presented approach reschedules the pending jobs in an event-driven manner (i.e., every time a new repair job is required). The event-driven scheduling process ensures that jobs of high importance, with a high emergency level, are completed promptly. This approach to event-driven replanning will allow companies to plan for real-world scenarios with significantly fewer resources than are used in practice.

A more expressive behavioral logic for decision-theoretic planning

We examine the problem of compactly expressing models of non-Markovian reward decision processes (NMRDP). In the field of decision-theoretic planning NMRDPs are used whenever the agent’s reward is determined by the history of visited states. Two different propositional linear temporal logics can be used to describe execution histories that are rewarding. Called PLTL and