This paper proposes a biologically-inspired collective behaviour for a team of co-operating robots. A modification of the subsumption architecture is proposed for implementing the control of the individual robots. The paper proposes a new approach of fuzzy logic technique to enable the resolution of conflicts between contradictory behaviours within each robot. Furthermore, a novel adaptive knowledge-based software architecture has been developed that enables team of robot agents to achieve adaptive cooperative control. Simulated and real experiments have been conducted on team of mobile robots performing two proof-of-concept co-operative tasks: dynamic target tracking and box-pushing.