This paper presents a novel approach to the adaptive scheduling of tasks in resource-limited event-driven distributed control architectures. The approach is suited to those applications which, owing to a combination of limited resources and complexity, do not admit the use of timed programming models (e.g. complex behavioural-based mobile robot control systems). The distributed scheduling mechanism uses a measurement of the cumulative fuzzy entropy at local points within the control system to determine whether a change is sufficiently significant to be transmitted over the network. The selective transmission of events, whilst still maintaining acceptable behaviour of the control system, reduces the network bandwidth usage and acts as scheduling mechanism. Since the frequency of events within such a system controls both the bandwidth usage and computation, the reduction in event frequency can also be used to reduce the overall power consumption of the system. The effect of entropy-based scheduling is illustrated by the simulation of a collision avoidance behaviour for a laboratory mobile robot. The results illustrate the application of the approach at different points within the architecture and for varying levels of significant threshold