Dominique Snyers

Routing in telecommunications networks with ant-like agents

A simple mechanism is presented, based on ant-like agents, for routing and load balancing in telecommunications networks, following the initial works of Appleby and Stewart [1] and Schoonderwoerd et al. [32,33]. In the present work, agents are very similar to those proposed by Schoonderwoerd et al. [32,33], but are supplemented with the ability to perform more computations at switching nodes, which significantly improves the networks relaxation and its response to perturbations.

Adaptive Agent-Driven Routing and Load Balancing in Communication Networks

This paper presents an unified overview of a new family of distributed algortithms for routing and load balancing in dynamic communication networks. These new algorithms are described as an extension to the classical routing algorithms: they combine the ideas of online asynchronous distance vector routing with adaptive link state routing. Estimates of the current traffic condition and link costs are measured by sending routing agents in the network that mix with the regular information packets and keep track of the costs (e.g. delay) encountered during their journey. The routing tables are then regularly updated based on that information without any central control nor complete knowledge of the network topology. Two new algorithms are proposed here. The first one is based on round trip routing agents that update the routing tables by backtracking their way after having reached the destination. The second one relies on forward agents that update the routing tables directly as they move toward their destination. An efficient co-operative scheme is proposed to deal with asymmetric connections. All these methods are compared on a simulated network with various traffic loads; the robustness of the new algorithms to network changes is proved on various dynamic scenarii.

Three-dimensional architectures grown by simple ‘stigmergic’ agents

A simple model of multi-agent three-dimensional construction is presented. The properties of this model are investigated. Based on these properties, a fitness function is defined to characterize the structured patterns that can be generated by the model. The fitness function assigns a value to each pattern. The choice of the fitness function is validated by the fact that human observers tend to view patterns with high (resp. low) fitness as structured (resp. unstructured). A genetic algorithm based on this fitness function is used to explore the space of possible patterns. The genetic algorithm is able to make use of sub-modules of existing patterns and recombine them to produce novel patterns, but strong epistatic interactions among genes make the fitness landscape rugged and prevent more complex patterns from being produced.

A Stochastic Heuristic for Visualising Graph Clusters in a Bi-DimensionalSpace Prior to Partitioning

This paper presents a new stochastic heuristic to reveal some structures inherent in large graphs, by displaying spatially separate clusters of highly connected vertex subsets on a two-dimensional grid. The algorithm employed is inspired by a biological model of ant behavior; it proceeds by local optimisations, and requires neither global criteria, nor any a priori knowledge of the graph. It is presented here as a preliminary phase in a recent approach to graph partitioning problems: transforming the combinatorial problem (minimising edge cuts) into one of clustering by constructing some bijective mapping between the graph vertices and points in some geometric space. After reviewing different embeddings proposed in the literature, we define a dissimilarity coefficient on the vertex set which translates the graphs interesting structural properties into distances on the grid, and incorporate it into the clustering heuristic. The heuristics performance on a well-known class of pseudo-random graphs is assessed according to several metric and combinatorial criteria.

Image processing optimization by genetic algorithm with a new coding scheme

Abstract An original coding scheme is introduced to take advantage of the two-dimensional structural information of images within the genetic algorithm framework. Results are presented showing that this new technique outperforms classical optimization methods for the optimization of 32 × 32 and 128 × 128 holograms.

Adaptive agent driven routing in communication networks: comparison with a classical approach

This paper follows an earlier publication in the Advances in Complex Systems journal (Heusse et al., 1998) where we presented a new algorithm based on collaborative agents for routing in communication networks. In this document, we shall investigate its load-balancing capability. This capability is required as a first step to achieve quality of delivery and service. We also compare our new approach to the classical ones and discuss their respective benefits.

Lotka Volterra Coevolution at the Edge of Chaos

In this paper, we study the coevolution of species by combining a theoretical approach with a computer simulation in order to show how a discrete distribution of viable species emerges. Coevolution is modelled as a replicator system which, with an additional diffusion term representing the mutation, leads to a Schrodinger equation. This system dynamics can be interpreted as a survival race between species on a multimodal sinking and drifting landscape whose modes correspond to the eigen modes of the Schrodinger equation. This coevolution dynamics is further illustrated by a simulation based on a continuous phenotypic model due to Kaneko in which the interactions between species are interpreted through a Lotka-Volterra model. This simulated coevolution is seen to converge to viable species associated with a dynamics at the edge of chaos (i.e with a null Lyapounov exponent). The transition from such a viable species to another results from some kind of tunnel effects characteristic of the punctuated equilibrium classically observed in biology in which rapid changes in the species distribution follow long plateaus of stable distribution.

Clique Partitioning Problem and Genetic Algorithms

In this paper we show how critical coding can be for the Clique Partitioning Problem with Genetic Algorithm (GA). Three chromosomic coding techniques are compared. We improve the classical linear coding with a new label renumbering technique and propose a new tree structured coding. We also show the limitation of an hybrid approach that combines GA with dynamic programming. Experimental results are presented for 30 and 150 vertex graphs.

Mimicry and Coevolution of Hedonic Agents

A hierarchy of cognitive agents is presented here in order to insist on the importance of interactions within a society of self teaching agents. In particular, this paper focuses on mimicry as the interaction mechanism of choice for hedonic agents. The dynamics of this mimetic type of interaction is studied both theoretically and by simulation. It leads to the coevolutionary dynamics best explained under the metaphor of an agent race on a landscape of sinking icebergs. A Darwinian selection mechanism is also included in the simulation model and the agents adapt their strategies by learning during one generation without being able to pass their findings directly to the next generation.