Yann Chevaleyre

Recent advances on multi-agent patrolling

Patrolling is a complex multi-agent task, which usually requires agents to coordinate their decision-making in order to achieve optimal performance of the group as a whole. In previous work, many patrolling strategies were developed, based on different approaches: heuristic agents, negotiation mechanisms, reinforcement learning techniques, techniques based on graph-theory and others. In this paper, we complement these studies by comparing all the approaches developed so far for this domain, using patrolling problem instances that were not dealt with before (i.e. new map topologies). The final results constitute a valuable benchmark for this domain, as well as a survey of the strategies developed so far for this task.

A Theoretical Analysis of Multi-Agent Patrolling Strategies

A group of agents can be used to perform patrolling tasks in a variety of domains ranging from computer network administration to computer wargame simulations. The quality of the strategies used for patrolling may be evaluated using different measures. Informally, a good strategy is one that minimizes the time lag between two passages to the same place and for all places. Recently, many different architectures of multi-agent systems have been proposed and evaluated on the patrolling problem [1]. In particular, different types of agents (reactive vs. cognitive), agent communication (allowed vs. forbidden), coordination scheme (central and explicit vs. emergent), agent perception (local vs. global), and decisionmaking mechanism (random selection vs. goal-oriented selection) were proposed. This paper proposes a theoretical analysis of the patrolling problem addressing the following issues : Are there efficient algorithms generating near-optimal strategies ? Are patrolling strategies based on partitioning the territory good?

Simple negotiation schemes for agents with simple preferences: sufficiency, necessity and maximality

We investigate the properties of an abstract negotiation framework where agents autonomously negotiate over allocations of indivisible resources. In this framework, reaching an allocation that is optimal may require very complex multilateral deals. Therefore, we are interested in identifying classes of valuation functions such that any negotiation conducted by means of deals involving only a single resource at a time is bound to converge to an optimal allocation whenever all agents model their preferences using these functions. In the case of negotiation with monetary side payments amongst self-interested but myopic agents, the class of modular valuation functions turns out to be such a class. That is, modularity is a sufficient condition for convergence in this framework. We also show that modularity is not a necessary condition. Indeed, there can be no condition on individual valuation functions that would be both necessary and sufficient in this sense. Evaluating conditions formulated with respect to the whole profile of valuation functions used by the agents in the system would be possible in theory, but turns out to be computationally intractable in practice. Our main result shows that the class of modular functions is maximal in the sense that no strictly larger class of valuation functions would still guarantee an optimal outcome of negotiation, even when we permit more general bilateral deals. We also establish similar results in the context of negotiation without side payments.

Learning Ordinal Preferences on Multiattribute Domains: The Case of CP-nets

A recurrent issue in decision making is to extract a preference structure by observing the user’s behavior in different situations. In this paper, we investigate the problem of learning ordinal preference orderings over discrete multiattribute, or combinatorial, domains. Specifically, we focus on the learnability issue of conditional preference networks, or CP-nets, that have recently emerged as a popular graphical language for representing ordinal preferences in a concise and intuitive manner. This paper provides results in both passive and active learning. In the passive setting, the learner aims at finding a CP-net compatible with a supplied set of examples, while in the active setting the learner searches for the cheapest interaction policy with the user for acquiring the target CP-net.

New candidates welcome! Possible winners with respect to the addition of new candidates☆

In voting contexts, some new candidates may show up in the course of the process. In this case, we may want to determine which of the initial candidates are possible winners, given that a fixed number k of new candidates will be added. We give a computational study of this problem, focusing on scoring rules, and we provide a formal comparison with related problems such as control via adding candidates or cloning.

How equitable is rational negotiation

Notions of fairness have recently received increased attention in the context of resource allocation problems, pushed by diverse applications where not only pure utilitarian efficiency is sought. In this paper, we study a framework where allocations of goods result from distributed negotiation conducted by autonomous agents implementing very simple deals. Assuming that these agents are strictly self-interested, we investigate how equitable the outcomes of such negotiation processes are. We first discuss a number of methodological issues raised by this study, pertaining in particular to the design of suitable payment functions as a means of distributing the social surplus generated by a deal amongst the participating agents. By running different experiments, we finally identify conditions favouring equitable outcomes.

Optimizing the Placement of Evacuation Signs on Road Network with Time and Casualties in Case of a Tsunami

In recent years, the number of people affected by natural disasters and in particular tsunamis has been increasing. Artificial Intelligence and Operation Research approaches to simulate crowd evacuation and make cities ready for Tsunamis are critical interest. Given an extremely simple model of human behavior, i.e. a memory-less stochastic agent, we address the problem of optimizing the placement of Tsunami evacuation signs with respect to evacuation time and casualties. Moreover, we formalize this optimization problem as a Mixed Integer Linear Programming (MILP) problem and we run some experiments with a MILP solver on two scenarios of early warning evacuation for the road network of Nhatrang city in Vietnam.

Welfare engineering in practice: on the variety of multiagent resource allocation problems

Many problems studied in the multiagent systems community can be considered instances of an abstract multiagent resource allocation problem. In this problem, which is now better understood theoretically, the goal is to satisfy a criterion of global optimality (formulated in terms of a suitable notion of social welfare), given that the agents sharing a set of resources follow a local rationality criterion reflecting their individual preferences. In this paper, we first show that this simple decentralised resource allocation framework allows us to model a wide variety of applications. These applications thereby benefit from all the theoretical results concerning the framework. We then draw up a list of criteria which can guide the application designer working within the framework and illustrate the relevance of our approach by discussing several applications in view of this list of design criteria.

Tractable negotiation in tree-structured domains

Multiagent resource allocation is a timely and exciting area of research at the interface of Computer Science and Economics. One of the main challenges in this area is the high complexity of negotiation. In particular, the complexity of the task of identifying rational deals, i.e. deals that are beneficial for all participants, often hinders the successful transfer of theoretical results to practical applications. To address this issue, we propose several protocols designed to tame the complexity of negotiation by exploiting structural properties of the utility functions used by agents to model their preferences over alternative bundles of resources. In particular, we consider domains where utility functions are k-additive (that is, synergies between different resources are restricted to bundles of at most k items) and tree-structured in the sense that the bundles for which there are synergies do not overlap. We show how protocols exploiting these properties can allow for drastically simplified negotiation processes.

Compilation and communication protocols for voting rules with a dynamic set of candidates

We address the problem of designing communication protocols for voting rules when the set of candidates can evolve via the addition of new candidates. We show that the necessary amount of communication that must be transmitted between the voters and the central authority depends on the amount of space devoted to the storage of the votes over the initial set of candidates. This calls for a bicriteria evaluation of protocols. We consider a few usual voting rules, and three types of storage functions: full storage, where the full votes on the initial set of voters are stored; null storage, where nothing is stored; and anonymous storage, which lies in-between. For some of these pairs (voting rule, type of storage) we design protocols and show that they are asymptotically optimal by determining the communication complexity of the rule under the storage function considered.