WSN-Based Wildlife Localization Framework in Dense Forests through Optimization Techniques

Abstract

Wildlife in forests can be threatened by different phenomena like deforestation and changes in land use. It introduces the need to track the diverse terrestrial species of animals to understand their moving patterns and distributions. This allows conservationists to evaluate which species need immediate policies to guarantee survival and how the different land uses impact the ecosystem dynamics. The tracking can be achieved by using collars installed in the animals' necks, namely End Nodes (ENs), which are GPS-based in most cases. However, this technology is energy-consuming and is constrained to the outdoors, with an adequate line of sight with the satellite network. These constraints, which are not always met in forests, motivate the need to develop Wireless Sensor Network(WSN)-based localization solutions where different Low Power Wide Area Network (LPWAN) protocols can be used. These protocols rely on the measurements of the Received Signal Strength Indicator (RSSI), the Time of Flight (ToF), and propagation models to determine the distance between the EN and a set of Anchor Nodes (ANs) with fixed and previously known positions, and then, apply trilateration techniques to estimate the position of the animal. However, the existing approaches might have significant errors due to multipath and shadow fading caused by dense canopies. To address these limitations, this paper proposes a framework to improve the localization accuracy in dense forests using a three-step strategy. First, we provide an optimization setup to adequately select the ANs positions, increasing the redundancy of trilateration and coverage. Then, we propose an optimized method to determine the distance between the EN and the ANs based on bias and variance minimization using the RSSI and ToF. Finally, the optimized redundancy and coverage setup and the optimized distance estimation are used to improve the localization using a scoring method for ANs with the most reliable distance estimations based on residual errors. Numerical studies show that our framework outperforms the state-of-the-art strategies regarding trilateration capacity, distance accuracy, and localization errors.