This paper presents the first stage of research into a multi-agent complex environment, called “FlatLand” aiming at emerging complex and adaptive obstacle-avoidance and targetachievement behaviors by use of a variety of learning mechanisms. The presentation includes a detailed description of the FlatLand simulated world, the learning mechanisms used as well as an efficient method for comparing the mechanisms’ performance, robustness and required computational effort.
|Title of host publication||11th IEEE Mediterranean Conference on Control and Automation (MED'03)|
|Number of pages||6|
|Publication status||Published - 1 Jun 2003|
- genetic algorithms
- machine learning
- simulated worlds