Bi-Criteria Optimization for Balancing Communication Connectivity and Distance Based on β-DQN in UAV Trajectory

Unmanned Aerial Vehicles (UAVs) offer significant advantages in maritime search and rescue by covering vast ocean areas rapidly. However, the complex maritime environment and limited communication coverage often necessitate multi-hop communication to maintain connectivity with the Ground Control Station (GCS). Optimizing UAV flight paths to minimize communication interruptions and reduce flight distance is crucial for conserving energy and time. This paper introduces the β-DQN algorithm, an enhancement of the traditional DQN algorithm, designed to optimize both communication connectivity and flight distance. The reward structure is divided into two parts: one for actions before reaching the destination, incorporating decay for faster convergence, and another for achieving the optimal path. This approach innovatively introduces the communication interruption penalty index β, modeled as a piecewise function that accounts for both communication penalties and distance-related rewards. Simulations show that β-DQN converges faster, with shorter paths and better connectivity than DQN. This reward design approach can also be applied to other reinforcement learning algorithms for improved bi-criteria path planning.