Papers by Mathijs de Weerdt
Journal of Artificial Intelligence Research, 2012
We present two new and efficient algorithms for computing all-pairs shortest paths. The algorithm... more We present two new and efficient algorithms for computing all-pairs shortest paths. The algorithms operate on directed graphs with real (possibly negative) weights. They make use of directed path consistency along a vertex ordering d. Both algorithms run in O(n^2 w_d) time, where w_d is the graph width induced by this vertex ordering. For graphs of constant treewidth, this yields O(n^2) time, which is optimal. On chordal graphs, the algorithms run in O(nm) time. In addition, we present a variant that exploits graph separators to arrive at a run time of O(n w_d^2 + n^2 s_d) on general graphs, where s_d
Joint Proceedings of the Workshop on AI Problems and Approaches for Intelligent Environments and Workshop on Semantic Cities - AIIP '13, 2013
En-route charging stations allow electric vehicles to greatly extend their range. However, as a f... more En-route charging stations allow electric vehicles to greatly extend their range. However, as a full charge takes a considerable amount of time, there may be significant waiting times at peak hours. To address this problem, we propose a novel navigation system, which communicates its intentions (i.e., routing policies) to other drivers. Using these intentions, our system accurately predicts congestion at charging stations and suggests the most efficient route to its user. We achieve this by extending existing time-dependent stochastic routing algorithms to include the battery's state of charge and charging stations. Furthermore, we describe a novel technique for combining historical information with agent intentions to predict the queues at charging stations. Through simulations we show that our system leads to a significant increase in utility compared to existing approaches that do not explicitly model waiting times or use intentions, in some cases reducing waiting times by over 80% and achieving near-optimal overall journey times. * This work was supported by the ORCHID (orchid.ac.uk) and iDEaS projects (ideasproject.info).
Proceedings of the fourth international joint conference on Autonomous agents and multiagent systems - AAMAS '05, 2005
Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics), 2008
Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics), 2009
Studies in Computational Intelligence, 2010
This paper presents a multi-player multi-issue negotiation model to solve a resource allocation p... more This paper presents a multi-player multi-issue negotiation model to solve a resource allocation problem. We design a multilateral negotiation protocol, by which rational players bid sequentially in consecutive rounds till a deadline. Every player's bid is a combination of all resource allocations for himself. In this framework, we perform a thorough theoretical analysis of the negotiation with complete information, which is a preliminary for the more complex incomplete information case. We show that, under a complete information setting, we can derive the negotiation strategies that form a subgame perfect equilibrium outcome. We also show that when a discount factor exists, an agreement will be reached immediately at the end of the first negotiation round. By making trade-offs between issues, the utility that every player gets in the equilibrium outcome is maximized and the solution is Pareto optimal.
Lecture Notes in Business Information Processing, 2010
We consider online mechanism design without money, where agents are allowed to trade items with o... more We consider online mechanism design without money, where agents are allowed to trade items with other agents, in an attempt to improve their own allocation. In an off-line context, this problem is known as the House Allocation Problem (HAP). We extend HAP to an online problem and call it the Online House Allocation Problem (OHAP). In OHAP, agents can choose when to arrive and depart over time and are allowed to be indifferent between items. Subsequently, we present our Agent Shifting Algorithm (ASA) for OHAP. A mechanism that uses ASA as its allocation rule is shown to be strategy-proof, individually rational and Pareto optimal. Moreover, we argue that any mechanism that obtains an outcome in OHAP that cannot be obtained by using ASA fails to be strategy-proof or is not Pareto optimal.
Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics), 2010
Demand responsive transportation has the potential to provide efficient public door-to-door trans... more Demand responsive transportation has the potential to provide efficient public door-to-door transport with a high quality. In currently implemented systems in the Netherlands, however, we observe a decrease in the quality of service (QoS), expressed in longer travel times for the customers. Currently, generally one transport company is responsible for transporting all customers located in a specified geographic zone. In general it is known that when multiple companies compete on costs, the price for customers decreases. In this paper, we investigate whether a similar result can be achieved when competing on quality instead. To arrive at some first conclusions, we set up a multiagent environment to simulate the assignment of rides to companies through an auction on QoS, and the insertion of allocated rides in the companies' schedules using online optimization. Our results reveal that this setup improves the quality of the service offered to the customers at moderately higher costs.
Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics), 2010
IJCAI International Joint Conference on Artificial Intelligence, 2011
We advocate the use of an explicit time representation in syntactic pattern recognition because i... more We advocate the use of an explicit time representation in syntactic pattern recognition because it can result in more succinct models and easier learning problems. We apply this approach to the real-world problem of learning models for the driving behavior of truck drivers. We discretize the values of onboard sensors into simple events. Instead of the common syntactic pattern recognition approach of sampling the signal values at a fixed rate, we model the time constraints using timed models. We learn these models using the RTI+ algorithm from grammatical inference, and show how to use computational mechanics and a form of semi-supervised classification to construct a real-time automaton classifier for driving behavior. Promising results are shown using this new approach.
2011 International Conference on Networking, Sensing and Control, ICNSC 2011, 2011
In most real-world settings, a transportation plan requires modifications during execution. A tho... more In most real-world settings, a transportation plan requires modifications during execution. A thorough evaluation of transportation planning methods thus requires testing and comparison in a dynamic environment. We give conditions on a simulation environment that follow from this requirement, and propose a multi-agent simulator meeting these conditions. In addition, we propose a new measure that captures robustness in such dynamic settings. The multi-agent simulator and the robustness measure are then used to compare three different transportation methods (two multi-agent planners and one online optimization approach) in settings with release time uncertainty and truck breakdown incidents.
IEEE/WIC International Conference on Intelligent Agent Technology, 2003. IAT 2003.
The aim of this paper is to combine standard planning and replanning methods into a rigorous unif... more The aim of this paper is to combine standard planning and replanning methods into a rigorous unifying framework, extending an existing logic-based approach to resource-based planning. In this Action Resource Framework (ARF), actions and resources are the primitive concepts. Actions consume and produce resources. Plans are structured objects composed of actions and resource schemes and an explicit dependency function specifying their interrelationships. Previous plans can be used both for creating new plans and for modifying plans. Since we are often interested in reusing only a part of these previous plans, we extend the Action Resource Formalism with incomplete plans. To eciently represent such incomplete plans we use the notion of a gap. We maintain a library of incomplete plans (with gaps), and we present operators to insert plan parts from this plan library into the current plan. We prove this set of plan operators to be complete. Generalizing the renement planning approach, we present a template algorithm for both planning and replanning using the ARF and its plan library and plan operators. Finally, we show that existing (re)planning methods and heuristics nicely t into this framework.
Proceedings of the fourth international joint conference on Autonomous agents and multiagent systems - AAMAS '05, 2005
We present a novel approach to multiagent planning for self-interested agents. The main idea behi... more We present a novel approach to multiagent planning for self-interested agents. The main idea behind our approach is that multiagent planning systems should be built upon (single-agent) plan repair systems. In our system agents can exchange goals and subgoals through an auction, using their own (planning) heuristics and utility functions to determine when to auction and what to bid. Some experimental results for a logistics domain show that this system can be used to support the coordination of self-interested agents.
Proceedings of the fourth international joint conference on Autonomous agents and multiagent systems - AAMAS '05, 2005
We discuss the application of Model Based Diagnosis in agent-based planning. We model a plan as a... more We discuss the application of Model Based Diagnosis in agent-based planning. We model a plan as a system to be diagnosed and assume that agents can monitor the execution of the plan by making partial observations during plan execution. These observations are used by the agents to explain plan deviations (errors) by qualifying some action instances as behaving abnormally. We prefer those qualifications that are maximum informative, i.e. explain as much as possible. Since in a plan several instances of the same action might occur, an error occurring in one instance might be used to predict the occurrence of the same error in an action instance to be executed later on. To account for such correlations, we introduce causal rules to generate diagnoses from action instances qualified as abnormally and we introduce Pareto minimal causal diagnoses as the right extension of classical minimal diagnoses. Next, we consider the multi-agent perspective where each agent is responsible for a part of the total plan, we show how plan-diagnoses of these partial plans are related to diagnoses of the total plan and how global diagnoses can be obtained in a distributed way.
Intelligent Infrastructures, 2009
An important problem in transportation is how to ensure efficient operational route planning when... more An important problem in transportation is how to ensure efficient operational route planning when several vehicles share a common road infrastructure with limited capacity. Examples of such a problem are route planning for automated guided vehicles in a terminal and route planning for aircraft taxiing at airports. Maintaining efficiency in such transport planning scenarios can be difficult for at least two reasons. Firstly, when the infrastructure utilization approaches saturation, traffic jams and deadlocks may occur. Secondly, incidents where vehicles break down may seriously reduce the capacity of the infrastructure and thereby affect the efficiency of transportation. In this chapter we describe a new approach to deal with congestion as well as incidents using an intelligent infrastructure. In this approach, infrastructural resources (road sections, crossings) are capable of maintaining reservations of the use of that resource. Based on this infrastructure, we present an efficient, context-aware, operational transportation planning approach. Experimental results show that our context-aware planning approach outperforms a traditional planning technique and provides robustness in the face of incidents, at a level that allows application to real-world transportation problems.
Transportation Research Part C: Emerging Technologies, 2010
Experiments studying the behavior of agent based methods over varying levels of uncertainty in co... more Experiments studying the behavior of agent based methods over varying levels of uncertainty in comparison to traditional optimization methods are generally absent from the literature. In this paper we apply two structurally distinct solution approaches, an on-line optimization and an agent based approach, to a drayage problem with time windows under two types of uncertainty. Both solution approaches are able to respond to dynamic events. The on-line optimization approach utilizes a mixed integer program to obtain a feasible route at 30-second intervals. The second solution approach deploys agents that engage in auctions to satisfy their own objectives based on the information they perceive and maintain locally. Our results reveal that the agent-based system can out perform the online optimization when service time duration is highly uncertain. The on-line optimization approach, on the other hand, performs competitively with the agent-based system under conditions of job arrival uncertainty. When both moderate service time and job arrival uncertainties are combined, the agent system outperforms the on-line optimization; however, in the case of extremely high combined uncertainty, the on-line optimization outperforms the agent-based approach.
Information and Computation, 2011
We develop theory on the efficiency of identifying (learning) timed automata. In particular, we s... more We develop theory on the efficiency of identifying (learning) timed automata. In particular, we show that: i) deterministic timed automata cannot be identified efficiently in the limit from labeled data, and ii) that one-clock deterministic timed automata can be identified efficiently in the limit from labeled data. We prove these results based on the distinguishability of these classes of timed automata. More specifically, we prove that the languages of deterministic timed automata cannot, and that one-clock deterministic timed automata can be distinguished from each other using strings in length bounded by a polynomial. In addition, we provide an algorithm that identifies one-clock deterministic timed automata efficiently from labeled data. Our results have interesting consequences for the power of clocks that are interesting also out of the scope of the identification problem.
Autonomous Agents & Multiagent Systems/Agent Theories, Architectures, and Languages, 2010
When agents need to interact in order to solve some (possi- bly common) problem, resolving potent... more When agents need to interact in order to solve some (possi- bly common) problem, resolving potential conflicts before- hand is often preferred to coordination during execution. Agents may lose some flexibility, but their course of action will be more predictable and often also more ecient, ob- taining a socially optimal outcome instead of a local opti- mum. One way to
Proceedings of the AgentLink Workshop on Practical Reasoning Agents (FAPR-00)}, Sep 20, 2000
We introduce a computational framework, consisting of resources, skills, goals and services to re... more We introduce a computational framework, consisting of resources, skills, goals and services to represent the plans of individual agents and to develop models and algorithms for cooperation processes between a collection of agents. Keywords: Teamwork and cooperation, multiagent planning, distributed resource allocation.
Proceedings of the AAAI Spring Symposium on Distributed Plan and Schedule Management, 2006
Many different problems are called distributed or multiagent planning problems. Existing approach... more Many different problems are called distributed or multiagent planning problems. Existing approaches each deal with only a subset of these problems. Which properties are essential in determining a solution method for multiagent planning problems? We argue that mainly the facts that communication is limited, agents have private goals, and the frequency of the dependencies between agents have a significant impact on the solution method.
Uploads
Papers by Mathijs de Weerdt