The video demonstrates the use of max-sum to coordinate the sense/sleep cycles of energy constrained sensors deployed in an urban environment to perform a wide area surveillance task. Sensors observe vehicles (shown as white dots) that move over the road network between random start and end points. Areas sensed by active sensors are shown in red in the video. The simulation is based on the RoboCup Rescue Simulation environment. Sensors have no prior knowledge regarding the distribution of vehicles or the overlapping sensing fields of neighbouring sensors. Therefore, prior to coordination, sensors synchronise their schedules to acquire information about the observations that they share with their neighbours. After this calibration phase, the max-sum algorithm is used to optimise the sense/sleep schedules of the sensors. During the coordination phase, the video shows the communication graphs that agents use to exchange the max-sum messages. When the coordination phase is complete (i.e., after a fixed number of message exchange), the sensors use the computed schedule. The upper graph on the right shows the measure that the system seeks to minimize; the percentage of events missed (e.g., the number of vehicles that were not detected by any sensors). This measure is computed in parallel for different sense/sleep policy: the continuous policy where the sensors are always active (this violates the energy constraints of the sensors and thus represents an upper bound on the system performance), the coordinated policy where the schedule is computed by the max-sum algorithm, the random policy where each sensor randomly chooses which cycle to sense without any coordination with the other sensors, and the synchronised policy where all the sensors adopt the same schedule. The graphs show that the max-sum policy outperforms the other energy constrained policies and reduces both the percentage of missed events as well as the average time it takes to first detect an event.