UAV-assisted energy-aware data collection in robotic wireless sensor networks under battery prediction error

Abstract

Unmanned Aerial Vehicles (UAVs) have become an important component of modern wireless sensor and robotic networks due to their flexible mobility, rapid deployment capability, and ability to collect data from spatially distributed nodes. In Robotic Wireless Sensor Networks (RWSNs), UAVs can act as mobile data collectors instead of static data collectors. To reduce the communication burden of the network, the sensor nodes in the network can be considered in a clustered structure. In this clustered structure, selected Cluster Head (CH) robots are visited by the UAV to collect data. Different approaches exist in the relevant literature regarding the UAV's limited battery capacity in UAV-assisted data collection scenarios. However, most current studies assume that the available UAV battery capacity is accurately known. In practice, battery charge status estimation can contain significant errors due to temperature variations, dynamic current demand, battery aging, sensor noise, and outdoor operating conditions. Since real-life experiments and simulations in the literature report that such errors can reach 5-8% as a limiting value in realistic UAV missions, this thesis also adopts 5-8% battery estimation error scenarios for UAV visits and, consequently, for the effects on energy transmission of nonvisited CH robots.And this thesis investigates an energy-efficient data acquisition problem in a multi-clustered robot network, specifically involving an UAV with limited battery capacity. In a clustered network, each cluster has a CH robot responsible for assigning tasks to other robots and collecting data from them. Using an imperfect battery estimation method, the aim is to optimally plan which CH robots the UAV can visit within the UAV's battery limits. Nonvisited CH robots transmit their data to the UAV via multi-hop communication. The UAV needs to make a trajectory decision that minimizes the data forwarding cost for nonvisited CH robots. Generally, it is assumed that there is accurate battery information, which is unrealistic and does not guarantee task completion. This study investigates the impact of battery estimation errors on energy efficiency in UAV-assisted RWSNs. Analyzing configurations with 5, 7, 10, and 50 CH robots, this study evaluates how 5% and 8% battery estimation errors affect data collection performance depending on the UAV's route choices, and how this affects the total energy consumption of the leader robots. The results show that errors have a limited impact on small-scale networks, but lead to significant increases in energy consumption as the network grows. Battery prediction errors, particularly regarding battery levels, can significantly increase the number of nonvisited channels and thus the energy consumed, while at certain battery levels, both the Robust Energy-Aware Data Collection Strategy (READCS) and READCS with imperfect battery prediction errors can visit all channels and are unaffected by the errors. These results highlight that battery prediction accuracy is critical, especially in large-scale wireless sensor networks, and that high battery levels improve system reliability. Secondly, this thesis considers the UAV visitation problem using a fuzzy logic-based decision mechanism. Fuzzy logic considers various parameters in visitation decisions. Here, a perspective is adopted where the UAV is influenced by parameters such as battery level, distance to CH robots, return distance to the base station, and data freshness or we can say data urgency. A Mamdani-type fuzzy inference system is designed to evaluate these inputs through linguistic rules and generate a visitability score for each candidate cluster head robot. Then, the UAV selects the next CH robot considering both the fuzzy decision score and the availability of remaining battery. In this way, the proposed method provides a flexible multi-criteria decision structure for UAV data collection. Thirdly, in this thesis, we extend the UAV-assisted data collection model by considering a more realistic three-dimensional Data-Aware Hovering Constraint Robust Energy-Aware Data Collection (3D-DAHC-READCS). In this model, CH robots and the UAV are represented with three-dimensional coordinates, and therefore, the UAV's travel energy is modeled considering not only horizontal distances but also altitude differences. In addition, a data-aware model, a different version of this scenario, is adopted. It is assumed that each cluster head robot collects a different amount of data from the sensor nodes around it; this directly affects the UAV's hovering time during data collection. Therefore, the UAV battery constraint includes both 3D travel energy and data-aware hovering energy. This extension makes it possible to analyze how heterogeneous data volumes and terrain-aware placement collectively affect UAV visit decisions, nonvisited cluster head clusters, and network transmission energy.