Iterative Learning Optimization for UAV Path Planning with Avoidance Zones

Effective path planning of unmanned aerial vehicles (UAVs) operating under avoidance zones is one of the critical capabilities to guarantee mission success. To obtain an optimized solution within reasonable computational time, an iterative learning optimization method (ILOM) is proposed to solve the UAV path planning problem with guaranteed computational performance in terms of convergence and objective value. First, the UAV path planning problem is formulated as a quadratically constrained quadratic programming (QCQP) problem. Next, a method combining matrix decomposition and iterative optimization is developed to solve QCQPs. However, computational performance is influenced by the algorithmic parameters involved in the iterative optimization method. Considering the implicit relationship between the algorithmic parameters and computational performance, convolutional neural network is applied to optimally select parameters in the iterative method instead of determining them from experience. Finally, the proposed ILOM is implemented in simulation examples to validate improved computational efficiency.