This paper discusses the key features of autonomous heterogeneous teams. Autonomy in this sense entails both the ability to overcome unforeseen circumstances and failures as well as the capability of enhancing performance over time. The paper presents a framework that provides a comprehensive solution for handling faults and dynamically configuring the team in response to varying situations, while improving its performance over time, on both individual and team levels, within a modular and expandable architecture. The performance of the framework is studied in a set of simulated foraging scenarios, under conditions of both software and hardware failure.