Vector Field based Control of Quadrotor UAVs for Wildfire Boundary Monitoring

Abstract

Accurate real-time information about an ongoing wildfire event is important for realizing effective and safe wildfire fighting. This paper is intended to solve the problem of guiding Unmanned Air Vehicles (UAVs) equipped with onboard cameras to monitor dynamic wildfire boundaries. According to whether the prior knowledge of the wildfire boundary is available or not, we propose a model-based vector field and a model-free vector field for UAV guidance. By describing the wildfire boundary with a zero level set function, the propagation of the wildfire boundary is modeled with the Hamilton-Jacobi equation. If the prior knowledge of the boundary is available, the typical radial basis function thin-plate spline is adopted to approximate the wildfire boundary and predicts its propagation. Then a 3D analytical vector field is constructed for an implicit function representing the wildfire boundary. If only partial observation of the wildfire boundary within the UAV’s field of view is available, the horizontal error between the UAV and its sensed segment of wildfire boundary and the vertical error between the UAV and the desired altitude are utilized to construct a 3D distance error based vector field, directly. To guide the UAV to converge to and patrol along the advancing wildfire boundary, the complex nonlinear dynamics of the UAV is exploited with differential flatness and incorporated with the above mentioned vector fields to design a nonlinear geometric controller. Computer simulations have been conducted to evaluate the performance of the proposed 3D vector field based controllers with both synthetic and real data, and simulation results demonstrate that the proposed algorithms can be effective methods to monitor the advancing wildfire boundaries.