Representations and solutions for game-theoretic problems
Artificial Intelligence - Special issue on economic principles of multi-agent systems
A game-theoretic formulation of multi-agent resource allocation
AGENTS '00 Proceedings of the fourth international conference on Autonomous agents
| Hi-index | 0.00 |
Different problems from several areas like network routing, traffic control in urban environments or economic optimization problems seem to have a similar theoretical background, especially in case, that somebody looks on strategies to solve them. Due to a high amount of input data and the high complexity of most of these tasks, it usually is very difficult or even impossible to find optimal solutions. Humans may be able to solve those problems still faster and better by using their intuition and by their ability to learn from their experience. The authors intend to show that a new approach to let an individual find sufficient solutions in a human-like manner is to rely on incremental exploration and on the adaptation of processes. Such a solution is not necessarily optimal but although it may satisfy the needs of the individual or a group of them. The individuals' decisions on what activities have to be performed to reach a certain goal are derived from previously successfully applied activity-sequences in combination with a trial-and-error-approach. Like the brain of a growing child, our algorithm is able to become more intelligent during time, i.e. the kept activity-sequences may consider more conditions by getting longer when the individual gains more experience. That also includes, that partial mistakes are made, learnt and that sequences must be erased after they have been identified to be inefficient in comparison to other strategies. The Library Game, introduced in this article, is a model related to the above-cited problems. It provides an environment to analyze and to proof the practicability and efficiency of our suggested solution.