PSO and Kalman Filter-Based Node Motion Prediction for Data Collection from Ocean Wireless Sensors Network with UAV

Abstract

In this paper, we consider a wireless sensor network of nodes at the sea surface drifting due to wind and sea currents. In our scenario an Unmanned Aerial Vehicle (UAV) will be used to gather data from the sensor nodes. The goal is to find a flyable path which is optimal in terms of sensor node energy consumption, total channel throughput between the UAV and sensor nodes, flight time for the UAV and frequency of the node visits by the UAV. Finally, the path should also be optimal concerning node position estimation uncertainty. A Kalman Filter (KF) is used to estimate the nodes motions and Particle Swarm Optimization (PSO) is the method used to calculate the UAV path taking all of these objectives into account. The proposed node tracking aware path planning solution is compared to two other scenarios: One where the path planning is based on full knowledge of the node positions at all times, and one where path planning is based on the last known positions of the nodes.

Path planning for UAV harvesting information from dynamical wireless sensor nodes at sea

“…In the networks and systems mentioned above, the wireless sensor nodes were usually considered static as in [14] or with limited mobility as in [15]. It is also common to assume that these wireless nodes will periodically send their position information to a ground-based operation center and use them for data processing and to plan an optimal trajectory for the UAVs.…”

“…These assumptions were partly relaxed in [15] by using a simple model for node motions which applies a constant ocean surface velocity at the area of interest. There are several factors considered when searching for an optimal trajectory, such as the positions, the node's energy, the priority for a specific type of sensing data, and the elapsed time since the last visit by the UAV.…”

“…The research in [15] showed that the position estimation error for the nodes would get significantly high (e.g., up to 10 km after a 1.5 hours flight) when the UAVs get updates from the nodes only when they are communicable of each other. Adoption of Kalman Filter (KF) improved the position estimation during the periods the UAVs could not contact the nodes.…”

“…1) A realistic ocean states measurement model: Instead of a simple model with constant ocean current as in [15], an accurate ocean model, the Norkyst-800, with high geographical resolution and hourly updates, will be used.…”

“…We will assume that the position of a node can only be measured when the UAV is within the communication distance with the node. This was modelled in [15] as…”