To be held in conjunction with the 

  Eleventh International Conference on Autonomous and Multi-Agent Systems, Valencia, Spain

The number of novel applications of  multi-agent  systems  has  followed  an exponential trend over the last  few  years,  ranging  from  online  auction design, through multi-sensor networks, to scheduling of tasks  in  multi-actor systems. Multi-agent  systems  designed  for  all these  applications generally involve some form of very hard optimization problems that are substantially  different from problems traditionally dealt  with  in  other areas  (e.g.  industrial processes or scheduling  applications).  

More  specifically,  the  technical issues that multi-agent algorithm designers have to deal with include :

• Algorithms to compute solutions to mechanisms that deal with different stakeholders who  may  be  self  interested  or  may  have  different computation/communication capabilities from their peers.

• Techniques to manage and disseminate relevant information across different agents.

• Dealing with privacy concerns: solving complex  optimization  problems while leaking as little private information as possible

• Problems that require anytime algorithms.

• Approximate algorithms that need to provide  guarantees  on  the  quality of the solution.

• Mechanisms whose properties  can  be  significantly  affected  if  the solution computed is not the optimal one.

• Techniques to deal with optimizations that have to  be  repeated  with possibly only slight changes in the input data.

• Techniques to deal  with  situations  where  the  input  data  may  be uncertain or unreliable, requiring that the solution computed be robust to slight differences from the true values.

  Techniques to deal with agents that are tied to physical devices. This involves computation and communication constraints that need to  be considered in the coordination techniques, as well as the possibility of failures of the devices and communication links.

Invited Talk

Dr. Christian Blum. (Talk blurb: coming soon.)

Christian Blum is an associate professor in computer science at the Universitat Politècnica de Catalunya (UPC). His research focuses on developing new swarm intelligence and hybrid metaheuristics techniques for solving optimization and control tasks that arise in many important applications areas such as telecommunications, bio-informatics, neuroscience and robotics. Concerning swarm intelligence, he has used algorithms such as ant colony optimization (ACO) and particle swarm optimization (PSO) for solving challenging combinatorial optimization problems and for problem solving in distributed environments such as adhoc and sensor networks. Concerning hybrids metaheuristics, he has mainly worked on two different types of hybridization: the hybridization of metaheuristics based on the construction of solutions with concepts from branch and bound and on the developing of efficient large neighborhood search algorithms. Concerning applications, apart from classical operations research applications, Christian has also applied these techniques to real-world applications in wireless sensor networks (self-organised duty-cycling in sensor networks) and in the bio-medical and neuroscience field (automated reconstruction of dentritic and axonal trees, DNA sequencing, training of neural networks for medical pattern classification, ...).

Additional information about Christian can be found on his webpage: http://www.lsi.upc.edu/~cblum/.

Keywords

• Distributed Constraint Optimisation/Satisfaction
• Winner Determination Algorithms in Auctions
• Coalition Formation Algorithms
• Algorithms to compute Nash Equilibrium in games
• Optimisation under uncertainty
• Optimisation with incomplete or dynamic input data
• Algorithms for real-time applications

Workshop Chairs

Dra. Meritxell Vinyals, University of Verona, Italy