Energy-Aware UAV Data Collection in Robotic Wireless Sensor Networks Under Imperfect Battery Predictions

This study investigates an energy-efficient data collection problem involving a mobile sink, specifically an unmanned aerial vehicle (UAV) with constrained battery capacity, within a network of robots organized into multiple clusters. Within each cluster, a cluster head (CH) robot is responsible for task allocation to the other robots and for data collection from them. With imperfect battery estimation techniques, we aim to reduce the total energy consumption of UAV while minimizing the cost of data collection from CH robots by optimally visiting a subset of these robots. The UAV selects the subset of CH robots to visit by considering both the locations of all CH robots and their battery capacity. If the UAV is unable to reach all CH robots, the CH robots that are not visited will transmit their data to another CH robot for forwarding. Most schedulers assume exact battery knowledge, which is unrealistic and can compromise feasibility while raising network energy consumption. This study investigates the impact of battery prediction (estimation) errors on energy efficiency in UAV-assisted robotic wireless sensor networks (RWSN). Analyzing configurations with 5 CH, 7 CH, and 10 CH robots, the analysis evaluates how battery prediction errors of 5% and 8% affect the UAV’s data collection performance based on the UAV’s routing preferences, and in turn, their impact on total energy consumption of the CH robots. The results show that for small-scale networks, the errors have a limited impact, but lead to significant increases in energy consumption as the network grows. Battery prediction errors, in particular battery levels, can significantly increase the number of unvisited CHs and consequently the consumed energy, whereas in some battery levels, both Robust Energy Aware Data Collection Strategy (READCS) and READCS with imperfect battery prediction errors can visit all CHs and are not affected by the errors. These results reveal that battery prediction accuracy is critical, especially in large-scale RWSNs, and high battery levels increase system reliability. The study provides an important baseline for future adaptive algorithms to improve energy efficiency.