Conference Paper

Multi-objective Parallel Tabu Search.

DOI: 10.1007/978-3-540-30217-9_74 Conference: Parallel Problem Solving from Nature - PPSN VIII, 8th International Conference, Birmingham, UK, September 18-22, 2004, Proceedings
Source: DBLP

ABSTRACT This paper describes the implementation of a parallel Tabu Search algorithm for multi-objective continuous optimisation problems.
We compare our new algorithm with a leading multi-objective Genetic Algorithm and find it exhibits comparable performance
on standard benchmark problems. In addition, for certain problem types, we expect Tabu Search to outperform other algorithms
and present preliminary results from an aerodynamic shape optimisation problem. This is a real-world, highly constrained,
computationally demanding design problem which requires efficient optimisation algorithms that can be run on parallel computers:
with this approach optimisation algorithms are able to play a part in the design cycle.

1 Bookmark
  • Source
    [Show abstract] [Hide abstract]
    ABSTRACT: The requirement of mitigating the environmental impact of helicopters has been addressed globally within the FRIENDCOPTER Project in the sen-se that several aspects (such as noise abatement, vibration reduction, fuel consumption) have been studied in parallel with different approaches among which there is the active blade control. In the work at hand, attention is paid on two topics aimed at mitigating helicopter environmental impact through a morphing strategy: the aerodynamic optimization of a blade section who-se camber can be affected by an actuator and the design of a SMA based static twist concept, aimed at extending the helicopter flight envelope. The latter device is based on a SMA rod which is integrated in the spanwise di-rection within the blade structure at different positions. The actuator, when heated, transmits a torque couple which induces twist onto the blade. The twist variation due to the SMA device activation has been predicted by a FE approach (MSC/Marc software implemented with the SMA Brinson model). Finally, rotor performance in hover has been estimated with the actuator in power on and off positions, highlighting benefits coming from several spanwise integrations/distributions of the SMA device.
  • Source
    [Show abstract] [Hide abstract]
    ABSTRACT: This study imitates the gene-therapy process at the forefront of medicine and proposes an innovative evaluative crossover operator. The evaluative crossover integrates a geneevaluation method with a gene-therapy approach in the traditional NSGA-II for finding uniformly distributed Pareto-optimal front of multi-objective optimization problems. To further enhance the advantages of fast non-dominate sorting and diversity preservation in NSGA-II, the proposed gene-evaluation method partially evaluates the merit of different crossover genes by substituting them in better parent and then calculating the fitness variances. The gene-therapy approach incorporates with the evaluative crossover to cure the mating parents mutually with respect to their gene contribution in order to retain superior genomes in the evolutionary population. Some comprehensive investigations for parameter setting are performed on a benchmark problem. Especially, three parameters of the evaluative crossover with a reasonable set of values are analyzed to realize their evolutionary effect. The experimental results show that a 100% crossover percentage with 10% crossover rate and a random therapeutic coefficient can achieve the best performance for E-NSGA-II. The proposed algorithm is tested on nine unconstrained multi-objective optimization problems. The experimental results are compared with five existing algorithms. The results show that the proposed E-NSGA-II is able to converge the Pareto-optimal front of all test problems, even though other algorithms experiences difficulties in approaching the global optima on some functions. E-NSGA-II can also achieve better diversity qualities than others. The results of algorithm analysis also reveal that the proposed evaluative crossover can intuitively evaluate gene contributions that can guide E-NSGA-II to perform an efficient search by dynamically shifting emphasis to significant genome in the feasible space without abdicating any portion of the candidate schemata. Although the convergence metric on the disconnected problem POL is slightly worse than four algorithms, E-NSGA-II outperforms almost all other algorithms on the mean of the convergence metric in other test problems. For the diversity metric, E-NSGA-II also performs better than all other algorithms dramatically in all test problems except for NSGA-II on POL and KUR. In the future, the proposed E-NSGA-II should further develop the properties of a simple yet efficient evaluative crossover operator, a revised mutation operator and a parameter-less approach to deal with a wide spectrum of real world multi-objective problems.
    03/2010; , ISBN: 978-953-307-055-1
  • [Show abstract] [Hide abstract]
    ABSTRACT: In order to resolve disturbances, which can affect a network transport, a regulation process has to be performed by a real-time regulator who has a complex decision-making role including disturbance detection, practicable decisions elaboration and evaluation; and choice decision to adopt. The objective of our research task consists of providing to regulator quickly solutions in case of disturbances in the transport network by integrating cybercars. We propose a multi-agent model for modeling that problem and cooperative approach for optimizing it. The cooperative approach is based on multi-objective tabu search algorithm and multi-objective particle swarm optimization algorithm. Keywords-component; trafc regulation; cybercar; multi- agent model; cooperative approach; multi-objective tabu search; multi-objective particle swarm optimization; Pareto dominance; multi-objective optimization I. INTRODUCTION Trafc planning of a public transport network starts by the construction of vehicles passages schedules in each station. It is what we call a periodic timetable. Timetable planners may satisfy the quality of the proposed services to their passengers by respecting at most the theoretical schedules announced in planning process. But, in real-time, various chances can appear and disturb global function of network. These disturbances can be resulted from trafc conditions and they are inuenced by courses time. Disturbances engender the modication of vehicles passage time by stations. By consequent, they result lack of punctuality and non compliance with the intervals between vehicles. It becomes necessary to act and to engage regulation operations trying to limit the impact of disturbances. Process regulation compares between services quality objec- tives and reality, and it operates susceptible actions toll them, considering exploitation constraints. This process allows the regulator to choose operations to be undertaken in order to reduce disturbances. Regulator has to modify the initial theo- retical timetable and he generates actions considering problem diagnosis and objectives assigned (6). Cybercar, which is a road vehicle with driving capacities totally automated, can be used as an alternative mode which will play the role of an additional help to regulator. Aeet of Cybercars consti- tutes intelligent transportation system that completes public transportation system to improve quality transport services. In fact, cybercars integration in trafc regulation process realize several advantages, it reduces characteristics of urban circles through driving assistant and technologies automation. Also, it decreases urban pollution through the use of automated electric vehicles. It assures a good quality of service thanks to demand services transportation (4). Trafc regulation by integrating cybercars is a complex problem which needs to be modeling and optimized. In this context, we propose a multi- agent model for trafc regulation by integrating cybercars. We elaborate a cooperative approach for optimization based on multi-objective tabu search and multi-objective particle swarm by considering concurrently all the exploitations criteria. This paper is organized as follows: In Section II, we list related works in trafc regulation. In Section III, we present some basic concepts from multi-objective optimization since trafc regulation is a multi-objective problem and we focus in multi- objective tabu search optimization and multi-objective particle swarm optimization . In section IV, we present mathematical formulation for that problem. In section V, we describe our multi-agent model using cooperative approach for optimization process. In section VI, we present our experimental study and wenish by conclusion in section VII.


Available from
Sep 18, 2014