1: \begin{abstract}

2:  Maintaining the freshness of information in the Internet of Things

3: (IoT) is a critical yet challenging problem. In this paper, we study

4: cooperative data collection using multiple Unmanned Aerial Vehicles

5: (UAVs) with the objective of minimizing the total average Age of Information

6: (AoI). We consider various constraints of the UAVs, including kinematic,

7: energy, trajectory, and collision avoidance, in order to optimize

8: the data collection process.  Specifically, each UAV, which has  limited

9: on-board energy, takes off from its initial location and flies over

10: sensor nodes to collect update packets in cooperation with the other

11: UAVs. The UAVs must land at their final destinations with non-negative

12: residual energy after the specified time duration to ensure they have

13: enough energy to complete their missions. It is crucial to design

14: the trajectories of the UAVs and the transmission scheduling of the

15: sensor nodes to enhance information freshness. We model the multi-UAV

16: data collection problem as a Decentralized Partially Observable Markov

17: Decision Process (Dec-POMDP), as each UAV is unaware of the dynamics

18: of the environment and can only observe a part of the sensors. To

19: address the challenges of this problem, we propose a multi-agent \textcolor{black}{Deep

20: Reinforcement Learning (DRL)-}based algorithm with centralized learning

21: and decentralized execution. In addition to the reward shaping, we

22: use  action masks to filter out  invalid actions and ensure that the

23: constraints are met. \textcolor{black}{Simulation results  demonstrate

24: that the proposed algorithms can significantly reduce the total average

25: AoI compared to the baseline algorithms, and the use of the action

26: mask method can improve the convergence speed of the proposed algorithm.}

27: \end{abstract}

28: