Christelle Bloch

Using an evolutionary algorithm to optimize the broadcasting methods in mobile ad hoc networks

A mobile ad hoc network (MANET) is a collection of mobile nodes communicating through wireless connections without any prior network infrastructure. In such a network the broadcasting methods are widely used for sending safety messages and routing information. To transmit a broadcast message effectively in a wide and high mobility MANET (for instance in vehicular ad hoc network) is a hard task to achieve. An efficient communication algorithm must take into account several aspects like the neighborhood density, the size and shape of the network, the use of the channel. Probabilistic strategies are often used because they do not involve additional latency. Some solutions have been proposed to make their parameters vary dynamically. For instance, the retransmission probability increases when the number of neighbors decreases. But, the authors do not optimize parameters for various environments. This article aims at determining the best communication strategies for each node according to its neighborhood density. It describes a tool combining a network simulator (ns-2) and an evolutionary algorithm (EA). Five types of context are considered. For each of them, we tackle the best behavior for each node to determine the right input parameters. The proposed EA is first compared to three EAs found in the literature: two well-known EAs (NSGA-II and SPEA2) and a more recent one (DECMOSA-SQP). Then, it is applied to the MANET broadcasting problem.

A self-adaptive multiagent evolutionary algorithm for electrical machine design

This paper presents a self-adaptive algorithm that hybridises evolutionary and multiagent concepts. Each evolutionary individual is implemented as a simple agent capable of re-production and predation. The transitions between these two states depend on the agents local environment. Thus, no explicit global process is defined to select neither the mates nor the preys. The convergence of the algorithm emerges from the behaviour of the agents. This brings interesting properties, such as population size self-regulation. Two sets of experimental results are provided: a comparison with Saw-Tooth Algorithm and micro-GA using four classical functions and an optimisation of the efficiency and the weight of an electrical motor. Some possible evolutions and prospects are finally proposed.

Designing Smart Adaptive Flooding in MANET Using Evolutionary Algorithm

This paper deals with broadcasting warning / emergency messages in mobile ad hoc networks. Traditional broadcasting schemes tend to focus on usually high and homogeneous neighborhood densities environments. This paper presents a broadcasting protocol that locally and dynamically adapts its strategy to the neighborhood densities. The behavior of the protocol is tuned using various internal parameters. Multiple combinations of those parameters have been pre-computed as optimal solutions for a range of neighborhood densities, and the most relevant one is dynamically chosen depending on the locally perceived environment. The combinations were determined by coupling an evolutionary algorithm and a network simulator, using a statistically realistic radio-propagation model (Shadowing Pattern). This approach is compared with other probabilistic methods while broadcasting an emergency message in vehicular ad hoc networks with variable and heterogeneous vehicle densities. In such a context, it is expected from the network to enable each node to receive the warning message. The results show that our protocol covers the whole network, whereas other methods only have a probability of 0.57 to 0.9 to cover the entire network.

Permanent magnet motor multiobjective optimization using multiple runs of an evolutionary algorithm

This paper presents an original method of permanent magnet motor optimal design. The permanent magnet machines optimization must respect multiple constraints. Efficiency and weight have a large influence on the design. These two constraints can be found in several vehicular applications: propulsion motors, electrical fans for combustion engine, driving motors for ancillaries, driving motors for air-circuit fuel-cell compressor...Indeed, in all those embedded applications, the efficiency must be maximal to limit the energy consumption and the mass or the volume must be as low as possible. In this paper, the authors focus on an original multi-objective optimization algorithm well adapted to the previous problem. The method is based on multiplying runs of a new genetic algorithm specialized in broadly covering the solution space around target objectives. This algorithm is an improved variant of previously developed algorithms. The efficiency of these algorithms was proven by comparing with a deterministic algorithm (SQP) and a reference multi-objective genetic algorithm (NSGA-II). The presented algorithm is first validated on a study case from the literature: the dimensioning of a slotless permanent magnet machine. Then experimental results of the complete method applied on a permanent magnet motor are highlighted in a multi-objective point of view.

Adaptive multi-objective genetic algorithm using multi-pareto-ranking

This paper extends an elitist multi-objective evolutionary algorithm, named GAME, based on several Pareto fronts corresponding to various fitness definitions. An additional operator is defined to create an adaptive version of this algorithm, called aGAME. This new operator alternates different modes of exploration of the search space all through an aGAME execution. Mode switching is controlled according to the values of two performance indicators, in order to maintain a good compromise between the quality and diversity of the returned solutions. aGAME is compared with the previous version (GAME) and with the three best-ranking algorithms of the CEC 2009 competition, using seven bi-objective benchmarks and the rules of this competition. This experimental comparison shows that aGAME outperforms these four algorithms, which validates both the efficiency of the proposed dynamic adaptive operator and the algorithm performance.

Optimizing communications in vehicular ad hoc networks using evolutionary computation and simulation

Broadcasting efficiently in a Vehicular Ad hoc Network (VANET) is a hard task to achieve. An efficient communication algorithm must take into account several aspects such as the neighboring density, the size and shape of the network, the use of the channel, the priority level of the message. Some studies [6, 12, 13] have proposed new solutions of broadcasting on such a network, but it is quite hard to evaluate their performance in various contexts. In order to determine the best repeating situation for each node in the network according to its environment, we developed a tool combining a network simulator (NS2) and an evolutionary algorithm. In this paper, we study four types of context and we tackle the best behavior for each node to determine the right input parameters. These studies are necessary to develop efficient broadcast algorithms in VANET.

New Notation and Classification Scheme for Vehicle Routing Problems

Vehicle Routing Problems have been some of the most studied problems in combinatorial optimisation because they have many applications in transportation and supply chain. They are usually known as Vehicle Routing Problems or VRPs. The related literature is quite large and diverse both in terms of variants of the problems and in terms of solving approaches. To identify the different variants of routing problems, authors generally use initialisms, in which various prefixes and suffixes indicate the presence of different assumptions or constraints. But this identification based on initialisms is inefficient. For example, two variants of a problem may be identified by the same abbreviation, whereas different abbreviations may be assigned to the same problem. This paper proposes a new notation and a new formalism to identify and to classify instances of routing problems. This contribution aims at filling in the gaps of the current identification system. The goal is to allow everyone to position his work accurately in the literature, and to easily identify approaches and results comparable to his research. The proposed notation is inspired by the scheduling formalism. It has four fields (π /α /β /γ ), respectively describing the type and horizon of the problem, the system structure, resources and demands, constraints and objectives to be optimized. 26 papers from the literature chosen for their disparity are classified using this notation to illustrate its usefulness and a software tool is proposed to make its use easier.

Broadcasting Information in Variably Dense Environment Using Connectionless Data Exchange (CoLDE)

Our main goal is to develop a new solution to use multi-tier broadcast in order to deliver messages to all the devices in certain areas, whether they are connected to different networks, or not connected to any network. We use beacons (management frames) in the IEEE 802.11 protocol to send data from the access point to the clients without an association, and we use the probe request/response to exchange small amounts of data without being connected to the same network and without threatening the security of any of them.

Multi-Pareto-Ranking evolutionary algorithm

This paper proposes a new multi-objective genetic algorithm, called GAME, to solve constrained optimization problems. GAME uses an elitist archive, but it ranks the population in several Pareto fronts. Then, three types of fitness assignment methods are defined: the fitness of individuals depends on the front they belong to. The crowding distance is also used to preserve diversity. Selection is based on two steps: a Pareto front is first selected, before choosing an individual among the solutions it contains. The probability to choose a given front is computed using three parameters which are tuned using the design of experiments. The influence of the number of Pareto fronts is studied experimentally. Finally GAMEs performance is assessed and compared with three other algorithms according to the conditions of the CEC 2009 competition.

Tuning an evolutionary algorithm with taguchi methods and application to the dimensioning of an electrical motor

This paper presents an original method of permanent magnet motor optimal design developped by both Electrical Engineering and Computer Science laboratories. An Evolutionary Algorithm combining Genetic Algorithms and Multiagent Systems is used. This Genetic Multiagent System parameters are determined using a robust design method based on the Taguchi approach. The quality of the algorithm is evaluated considering the multiobjective quality of the solutions it delivers on a permanent magnet machine constrained optimization. Contradictory objectives as efficiency and weight have a large influence on the design of electrical machines. Performances of the resulting tuned up algorithm are compared with previous results from the authors.