Unmanned aerial vehicles (UAVs) have gained popularity due to their flexible mobility, on-demand deployment, and ability to establish line-of-sight wireless communication. However, existing UAV-assisted communication schemes often overlook the critical issue of collision avoidance during UAV flight. This paper proposes an interpretable UAV-assisted communication scheme that addresses this challenge through decomposition into two sub-problems. The first sub-problem involves constrained UAV coordinates and power allocation, solved using the Dueling Double DQN (D3QN) method. The second sub-problem deals with constrained UAV collision avoidance and trajectory optimization, addressed through the Monte Carlo tree search (MCTS) method. This approach ensures reliable and efficient UAV operation. To enhance the transparency and reliability of system decisions, a scalable explainable artificial intelligence (XAI) framework is proposed. The interpretability of the scheme generates explainable and trustworthy results, facilitating comprehension, validation, and control of UAV-assisted communication solutions. Extensive experiments demonstrate the superiority of the proposed algorithm in terms of performance and generalization compared to existing techniques. The proposed model improves the reliability, efficiency, and safety of UAV-assisted communication systems, offering a promising solution for future applications.