FreeWill+ is a framework that aims at integrating various animation techniques for controlling human-like characters. With heterogeneity and multi-layering as its main design principles, the system contains a society of complex interrelated animation components (actions), executing on various levels of abstraction and also varying in the level of built-in intelligence. They arrange themselves into ad-hoc created associations to perform tasks and fulfil goals. As both reactive and proactive modes of operation are present, and the communication between actions is crucial, the overall structure may be considered as a multiagent system. We will show how a relatively simple, multi-layered and multi-agent structure can cope with collision-free motion planning problem.