Most existing multi-unmanned aerial vehicle (multi-UAV) systems focus on fly path or energy consumption for task assignment, while little attention has been paid to the dynamic feature of the task, resulting in poor task completion ratio. The machine learning (ML) paradigm provides new methodologies for task assignment. However, ML methods are usually of heavy resource-consumption that cannot be directly applied in the UAV. In this paper, a digital twin (DT) assisted task assignment approach is proposed to improve the resource-intensive utilization and the efficiency of deep reinforcement learning (DRL) in multi-UAV system. The approach has a three-layer network structure which can dynamically assign tasks based on the task time constraints. Moreover, the approach is divided into two stages of initial task-assignment and task-reassignment. In the first stage, airship divides a task into multiple subtasks according to the shortest distance based on genetic algorithm and assigns them to UAVs. In the second stage, the DT can be leveraged to enable the airships to learn from the features of tasks and to generate the Q-value of the estimated value network of DRL for UAVs via pre-train of DT. The Q-value can be directly applied for deep Q-learning network (DQN) in the UAVs to reduce the training episode. Furthermore, the DQN is adopted to train taskreassignment strategy. Simulation results indicate that the DQN with DT can significantly reduce the training episode, improving 30% of the task completion ratio and 19% of the system energy efficiency compared with that of the baseline methods.

Ensuring that a vehicle can obtain its real location in a high-precision prebuilt map is one of the most important tasks of the Autonomous Internet of Vehicles (AIoV). In this work, we show that the resampling of the particle filter (PF) algorithm is optimized by using the prior information of particles that shift real localizations to improve vehicle localization accuracy without changing the existing PF process, i.e., the particle shift filter (PSF). The number of particles is critical to their convergence efficiency. We perform quantitative and qualitative analyses of how to improve particle localization accuracy while ensuring timeliness, without increasing the number of particles. Moreover, the cumulative error of the particles increases with time, and the localization accuracy and robustness decrease. Our findings show that the initial particle density is 159 particles/m3, and the cumulative variance of the PSF particles is improved by 27%, 29%, and 82% at the x-, y-, and z-axes, respectively, under the same conditions as the PF algorithm, while the calculation time only increases by 10.6%. Moreover, the cumulative mean error is reduced by 0.74 m, 0.88 m, and 0.68 m at the x-, y-, and z-axes, respectively, indicating that the localization error of the PSF changes less with time. All experiments were performed using open-source software and datasets.