The operations of autonomous IoT systems require the real-time analysis of complex signals acquired from the surrounding environment. This need contrasts with the limited computation capabilities of most IoT devices. Offloading computation to compute-capable devices such as edge servers can improve the response time of real-time systems, but also faces important challenges. In fact, the dynamics of the wireless channel connecting the device producing the data stream to the edge server, computation load of the latter and characteristics of the signals may determine frequent shift in the optimal computation location. This talk will summarize our work in adaptive and distributed computing for autonomous Unmanned Aerial Vehicles and healthcare IoT systems.