All Bids for One and One Does for All: Market-Driven Multi-agent Collaboration in Robot Soccer Domain

Abstract

Abstract. This work proposes a novel approach for introducing market-driven strategy to robot soccer domain in order to solve vital issues related to multiagent coordination. In robot soccer, two teams of robots compete with each other to win the match. For the benefit of the team, the robots should work collaboratively, whenever possible. Market-driven approach applies the basic properties of free market economy to a team of robots, to increase the profit of team as much as possible. This approach is based on the assumption that maximizing individual profit will approximate global profit maximization. In several works, this method was applied to some open issues in multi-agent systems like multi-robot exploration and coordination, but these implementations were limited. In this work, this approach is applied for the first time to the robot soccer domain, which is being a complex, dynamic and real-time event, one of the prominent topics of multi-agent research. In this paper, a market-driven collaborative task allocation algorithm for the robot soccer domain is proposed and experimental results are discussed.

Soccer without intelligence

“…In order to let the team develop some defensive skills, the team was trained against the BrianTeam team, the codes of which come with the TeamBots simulator. Finally, the robots were trained against a strong team named MarketTeam, which uses a market-driven role allocation algorithm and a similar potential field approach in which the coefficients of the field forces are trained by using Genetic Algorithms (GA) [19], [20]. Being a strong and offensive team, MarketTeam forced our team to learn some defensive behavior.…”

“…The algorithms were implemented and tested on the TeamBots mobile robot simulator environment [18]. Köse et al [19], [20] presented a task allocation algorithm based on free-market approach, in which the players evaluate their fitnesses to each role by calculating a cost function for accomplishing the main task of that role. These costs are exchanged among teammates and whichever player offers the lowest cost for a role is assigned to that role.…”

Multi-agent role allocation: issues, approaches, and multiple perspectives

“…Kim [80] Agents broadcast and use information about local state to determine roles Köse et al [88] Agents broadcast and use information about cost for executing roles Köse et al [89] Agents use reinforcement learning to discover role-taking policies Vail and Veloso [161] Agents broadcast and use information about local state to determine roles Wang et al [167] Minority-game inspired approach…”

Market-Driven Multi-Agent Collaboration in Robot Soccer Domain

“…As a result, every robot calculates its own bid for each action according to the following equations and broadcasts only these values (Kose et al, 2003;Frias-Martinez et al, 2004).…”

“…Domains like agricultural areas are simple, static and do not require fast task allocation, planning and coordination as in robot soccer (Kose et al, 2004). In order to provide a satisfactory solution to the task assignment and collaboration problem in robot soccer, several approaches have been implemented including static assignment (Kaplan, 2003), market based assignment (Kose et al, 2003) and reinforcement learning based extension to market based approach (Kose et al, 2004;Tatlidede et al, 2005). In this chapter, these approaches are compared and studied in detail.…”