This paper describes the autonomous behavior of block shaped agents existing in a spatially constrained environment. The difficulty of this problem is the determination of the behavior of a block agent under a complex environment dynamically produced mutual actions of many block agents. Moving obstacle avoidance problem causes this difficulty resulting from the autonomous motion of block agents. The objective of this research is to build an adaptive decision mechanism of the behavior for the autonomous block agents when they have their own tasks. Specifically, this paper shows one case study of the proposed mechanism for the problem of removing blocks from a container. It is assumed that the agents are block shaped, and the positions and dimensions of each block are different. Our approach adopted is the Classifier System based architecture. The results of simulation experiments are presented which indicate the possibility of this architecture in allowing block agents to adapt to dynamic environments.
