“…In multi-agent systems a straightforward approach is centralised workload distribution: a main controller collects all required data, such as robots' positions and sensor readings, then allocates scheduled tasks to the most suitable robots (Keshmiri and Payandeh, 2011;Liu and Kroll, 2012) in order to achieve the goal. Despite the simplicity of this approach and recent breakthroughs within indoor environments (D'Andrea, 2012), there are several well-known challenges related to the accuracy and availability of global information; precise information is not always possible to obtain, especially within large domains, for instance outdoor (Reggente and Lilienthal, 2009) or unstructured scenarios (Chitta et al, 2012). In addition, the robustness of such systems might be compromised when a problem or failure affects the central controller, which could lead to a halt in overall system operation.…”