傳統的無人機辨識和響應（跟蹤和避障）方法在面對環境不確定性方面存在著局限性，例如光照變化、視角變化和物體模糊性等。挑戰在於建立可靠的數學模型，將感知輸入與後續行動之間建立起來。複雜的人工智能模型，通常被稱為「黑盒子」，缺乏解釋性，使用戶難以信任其決策過程。此外，硬件限制和延遲問題也影響了即時的決策過程。本文介紹了一種基於可解釋、分散式和適應性人工智慧的無人機系統。運動控制模型使用參數化深度 Q 網絡算法進行訓練，而識別模型則使用了 You Only Look Once v4 Tiny（自適應 AI）方法。訓練後的模型部署在邊緣運算設備上，實現了分散式人工智慧。最後，使用模糊推理系統和 Local Interpretable Model-Agnostic Explanations 對模型的輸出進行合理的推論，從而實現了可解釋的人工智慧。這三個子系統根據它們在避障和跟蹤方面的性能進行評估。強化學習模型分別提供了短期風險指標為 -0.08 和 -1.65，以及長期風險指標為 0.01 和 0.12。與現有的 Deep Q-network 和 Deep Deterministic Policy Gradient 等方法相比，所提出的方法具有更優越的收斂速度。辨識模型使用了精確度、召回率、F1-score 和平均精確度等指標進行評估。

Traditional methods for the identification and response (tracking and avoidance) of Unmanned Aerial Vehicles (UAVs) are limited in their robustness against environmental uncertainties, such as changes in illumination, perspective variations, and object ambiguities. The challenge lies in establishing reliable mathematical models between sensory input and subsequent actions. Complex Artificial Intelligence (AI) models, often referred to as 'black boxes,' lack interpretability, making it hard for users to trust their decision-making process. Additionally, hardware limitations and latency issues hinder real-time decision-making. This paper introduces a UAV system based on Explainable, Distributed, and Adaptive AI. Motion control models are trained using the Parameterized Deep Q-network algorithm, while the identification of friend or foe is executed using the You Only Look Once v4 Tiny (Adaptive AI) method. The trained models are deployed on edge computing devices, manifesting Distributed AI. Lastly, Fuzzy Inference System and Local Interpretable Model-Agnostic Explanations are utilized to provide reasonable inference on the model's output, thereby achieving Explainable AI. The three subsystems are evaluated in terms of their performance in avoidance and tracking. The reinforcement learning model delivers short-term risk scores of -0.08 and -1.65, and long-term risk scores of 0.01 and 0.12, respectively. When compared with the existing methods like Deep Q-network and Deep Deterministic Policy Gradient, the proposed method exhibits superior convergence speed. The identification model is evaluated using metrics such as Precision, Recall, F1-score, and Average Precision (AP).

[1] Z. Y. Wang, D. Zhao, and Y. F. Cao, “Image Quality Enhancement with Applications to Unmanned Aerial Vehicle Obstacle Detection,” Aerospace, vol. 9, no. 12, Dec 2022
[2] S. C. Wu, R. Li, Y. J. Shi, and Q. S. Liu, “Vision-Based Target Detection and Tracking System for a Quadcopter,” Ieee Access, vol. 9, pp. 62043-62054, 2021.
[3] Y. M. Li, C. H. Fu, Z. Y. Huang, Y. Q. Zhang, and J. Pan, “Intermittent Contextual Learning for Keyfilter-Aware UAV Object Tracking Using Deep Convolutional Feature,” Ieee Transactions on Multimedia, vol. 23, pp. 810-822, 2021.
[4] K. F. Wan, X. G. Gao, Z. J. Hu, and G. F. Wu, “Robust Motion Control for UAV in Dynamic Uncertain Environments Using Deep Reinforcement Learning,” Remote Sensing, vol. 12, no. 4, Feb 2020.
[5] J. Zhao et al., “Deep Reinforcement Learning-Based End-to-End Control for UAV Dynamic Target Tracking,” Biomimetics, vol. 7, no. 4, Dec 2022.
[6] R. J. Ding, Y. D. Xu, F. F. Gao, and X. M. Shen, “Trajectory Design and Access Control for Air-Ground Coordinated Communications System With Multiagent Deep Reinforcement Learning,” Ieee Internet of Things Journal, vol. 9, no. 8, pp. 5785-5798, Apr 2022.
[7] A. Barredo Arrieta et al., “Explainable Artificial Intelligence (XAI): Concepts, taxonomies, opportunities and challenges toward responsible AI,” Information Fusion, vol. 58, pp. 82-115, 2020.
[8] F. Sado, C. K. Loo, W. S. Liew, M. Kerzel, and S. Wermter, “Explainable Goal-driven Agents and Robots - A Comprehensive Review,” Acm Computing Surveys, vol. 55, no. 10, Oct 2023, Art no. 211.
[9] P. McEnroe, S. Wang, and M. Liyanage, “A Survey on the Convergence of Edge Computing and AI for UAVs: Opportunities and Challenges,” Ieee Internet of Things Journal, vol. 9, no. 17, pp. 15435-15459, Sept 2022.
[10] G. P. Rong, Y. C. Xu, X. X. Tong, and H. J. Fan, “An edge-cloud collaborative computing platform for building AIoT applications efficiently,” Journal of Cloud Computing-Advances Systems and Applications, vol. 10, no. 1, Jul 2021, Art no. 36.
[11] B. Rubi, B. Morcego, and R. Perez, “Quadrotor Path Following and Reactive Obstacle Avoidance with Deep Reinforcement Learning,” Journal of Intelligent and Robotic Systems: Theory and Applications, vol. 103, no. 4, 2021.
[12] J. Ou, X. Guo, M. Zhu, and W. Lou, “Autonomous quadrotor obstacle avoidance based on dueling double deep recurrent Q-learning with monocular vision,” Neurocomputing, vol. 441, pp. 300-310, 2021.
[13] J. Hu, X. Yang, W. Wang, P. Wei, L. Ying, and Y. Liu, “Obstacle Avoidance for UAS in Continuous Action Space Using Deep Reinforcement Learning,” IEEE Access, vol. 10, pp. 90623-90634, 2022.
[14] L. He, N. Aouf, and B. F. Song, “Explainable Deep Reinforcement Learning for UAV autonomous path planning,” Aerospace Science and Technology, vol. 118, Nov 2021, Art no. 107052.
[15] X. H. Deng, Y. J. Liu, C. X. Zhu, and H. G. Zhang, “Air-Ground Surveillance Sensor Network based on edge computing for target tracking,” Computer Communications, vol. 166, pp. 254-261, Jan 2021, doi: 10.1016/j.comcom.2020.10.012.
[16] P. Wang, S. Huang, W. L. Leong, Z. Ma, S. Srigrarom, and R. S. H. Teo, “Low-cost camera based sense and avoid in unmanned aerial vehicles: Sensing and control methods,” Mechatronics, vol. 87, 2022.
[17] B. Li, Z.-p. Yang, D.-q. Chen, S.-y. Liang, and H. Ma, “Maneuvering target tracking of UAV based on MN-DDPG and transfer learning,” Defence Technology, vol. 17, no. 2, pp. 457-466, 2021.
[18] B. Li, Z. Gan, D. Chen, and D. S. Aleksandrovich, “Uav maneuvering target tracking in uncertain environments based on deep reinforcement learning and meta-learning,” Remote Sensing, vol. 12, no. 22, pp. 1-20, 2020.
[19] K. Wan, D. Wu, Y. Zhai, B. Li, X. Gao, and Z. Hu, “An Improved Approach towards Multi-Agent Pursuit–Evasion Game Decision-Making Using Deep Reinforcement Learning,” Entropy, vol. 23, no. 11, p. 1433, 2021.
[20] N. Abo Mosali, S. S. Shamsudin, O. Alfandi, R. Omar, and N. Al-Fadhali, “Twin Delayed Deep Deterministic Policy Gradient-Based Target Tracking for Unmanned Aerial Vehicle with Achievement Rewarding and Multistage Training,” IEEE Access, vol. 10, pp. 23545-23559, 2022.
[21] C. Liu and E.-J. van Kampen, “HER-PDQN: A Reinforcement Learning Approach for UAV Navigation with Hybrid Action Spaces and Sparse Rewards,” in AIAA Science and Technology Forum and Exposition, San Diego, CA, United states, 2022.
[22] H. X. Pham, H. M. La, D. Feil-Seifer, and L. V. Nguyen, “Autonomous UAV navigation using reinforcement learning,” arXiv, 2018.
[23] V. Mnih et al., “Human-level control through deep reinforcement learning,” Nature, vol. 518, no. 7540, pp. 529-533, 2015/02/01 2015.
[24] S. F. Abedin, M. S. Munir, N. H. Tran, Z. Han, and C. S. Hong, “Data Freshness and Energy-Efficient UAV Navigation Optimization: A Deep Reinforcement Learning Approach,” IEEE Transactions on Intelligent Transportation Systems, vol. 22, no. 9, pp. 5994-6006, 2021.
[25] Y. Yang, K. Zhang, D. Liu, and H. Song, “Autonomous UAV navigation in dynamic environments with double deep Q-networks,” in 39th AIAA/IEEE Digital Avionics Systems Conference, Virtual, San Antonio, TX, United states, 2020.
[26] X. Han, J. Wang, J. Xue, and Q. Zhang, “Intelligent Decision-Making for 3-Dimensional Dynamic Obstacle Avoidance of UAV Based on Deep Reinforcement Learning,” in 11th International Conference on Wireless Communications and Signal Processing, Xi'an, China, 2019.
[27] C. Yan, X. Xiang, and C. Wang, “Towards Real-Time Path Planning through Deep Reinforcement Learning for a UAV in Dynamic Environments,” Journal of Intelligent and Robotic Systems: Theory and Applications, vol. 98, no. 2, pp. 297-309, 2020.
[28] S. Zhang, Y. Li, and Q. Dong, “Autonomous navigation of UAV in multi-obstacle environments based on a Deep Reinforcement Learning approach,” Applied Soft Computing, vol. 115, 2022.
[29] Z. Cai and N. Vasconcelos, “Cascade R-CNN: High quality object detection and instance segmentation,” IEEE Transactions on Pattern Analysis and Machine Intelligence, vol. 43, no. 5, pp. 1483-1498, 2021.
[30] Y. Liu, F. Yang, and P. Hu, “Small-Object Detection in UAV-Captured Images via Multi-Branch Parallel Feature Pyramid Networks,” IEEE Access, vol. 8, pp. 145740-145750, 2020.
[31] D. Avola et al., “MS-faster R-CNN: Multi-stream backbone for improved faster R-CNN object detection and aerial tracking from UAV images,” Remote Sensing, vol. 13, no. 9, 2021.
[32] X. Liang, J. Zhang, L. Zhuo, Y. Li, and Q. Tian, “Small Object Detection in Unmanned Aerial Vehicle Images Using Feature Fusion and Scaling-Based Single Shot Detector with Spatial Context Analysis,” IEEE Transactions on Circuits and Systems for Video Technology, vol. 30, no. 6, pp. 1758-1770, 2020.
[33] M. Liu, X. Wang, A. Zhou, X. Fu, Y. Ma, and C. Piao, “Uav-yolo: Small object detection on unmanned aerial vehicle perspective,” Sensors (Switzerland), vol. 20, no. 8, 2020.
[34] J. M. Mendel and P. P. Bonissone, “Critical Thinking About Explainable AI (XAI) for Rule-Based Fuzzy Systems,” Ieee Transactions on Fuzzy Systems, vol. 29, no. 12, pp. 3579-3593, Dec 2021.
[35] B. Aldughayfiq, F. Ashfaq, N. Z. Jhanjhi, and M. Humayun, “Explainable AI for Retinoblastoma Diagnosis: Interpreting Deep Learning Models with LIME and SHAP,” Diagnostics, vol. 13, no. 11, Jun 2023, Art no. 1932.
[36] H. Y. Chen and C. H. Lee, “Vibration Signals Analysis by Explainable Artificial Intelligence (XAI) Approach: Application on Bearing Faults Diagnosis,” Ieee Access, vol. 8, pp. 134246-134256, 2020.
[37] M. Suffian, P. Graziani, J. M. Alonso, and A. Bogliolo, “FCE: Feedback Based Counterfactual Explanations for Explainable AI,” Ieee Access, vol. 10, pp. 72363-72372, 2022.
[38] H. M. Zhang and L. Hanzo, “Federated Learning Assisted Multi-UAV Networks,” Ieee Transactions on Vehicular Technology, vol. 69, no. 11, pp. 14104-14109, Nov 2020.
[39] H. L. Yang, J. Zhao, Z. H. Xiong, K. Y. Lam, S. M. Sun, and L. Xiao, “Privacy-Preserving Federated Learning for UAV-Enabled Networks: Learning-Based Joint Scheduling and Resource Management,” Ieee Journal on Selected Areas in Communications, vol. 39, no. 10, pp. 3144-3159, Oct 2021.
[40] Y. T. Wang, Z. Su, N. Zhang, and A. Benslimane, “Learning in the Air: Secure Federated Learning for UAV-Assisted Crowdsensing,” Ieee Transactions on Network Science and Engineering, vol. 8, no. 2, pp. 1055-1069, Apr-Jun 2021.
[41] P. Corke, Robotics, Vision and Control, 2 ed. (Springer Tracts in Advanced Robotics). 2017.
[42] J. Xiong et al., “Parametrized deep Q-networks learning: Reinforcement learning with discrete-continuous hybrid action space,” arXiv, 2018.
[43] W. Masson, P. Ranchod, and G. Konidaris, “Reinforcement learning with parameterized actions,” in 30th AAAI Conference on Artificial Intelligence, Phoenix, AZ, United states, 2016, pp. 1934-1940.
[44] M. Hausknecht and P. Stone, “Deep reinforcement learning in parameterized action space,” in 4th International Conference on Learning Representations, San Juan, Puerto rico, 2016.
[45] G. Xiong and L. Dong, “Vision Based Autonomous Tracking of UAVs Based on Reinforcement Learning,” in 2020 Chinese Automation Congress, Shanghai, China, 2020, pp. 2682-2686.
[46] S. Yang, Z. Meng, X. Chen, and R. Xie, “Real-time obstacle avoidance with deep reinforcement learning* Three-Dimensional Autonomous Obstacle Avoidance for UAV,” in 2019 International Conference on Robotics, Intelligent Control and Artificial Intelligence, Shanghai, China, 2019, pp. 324-329.
[47] C.-Y. Wang, A. Bochkovskiy, and H.-Y. M. Liao, “Scaled-yolov4: Scaling cross stage partial network,” in 2021 IEEE/CVF Conference on Computer Vision and Pattern Recognition, Virtual, Online, United states, 2021, pp. 13024-13033.
[48] L. Y. Lo, C. H. Yiu, Y. Tang, A. S. Yang, B. Y. Li, and C. Y. Wen, “Dynamic Object Tracking on Autonomous UAV System for Surveillance Applications,” Sensors, vol. 21, no. 23, Dec 2021.
[49] M. Alshehri, “Fuzzy Logic Based Explainable AI Approach for the Easy Calibration of AI Models in IoT Environments,” in 13th International Conference on Intelligent Human Computer Interaction, Kent, United states, 2022, vol. 13184 LNCS, in Lecture Notes in Computer Science, pp. 621-633.
[50] S. C. Y. Chan, S. Fishman, J. Canny, A. Korattikara, and S. Guadarrama, “Measuring the reliability of reinforcement learning algorithms,” arXiv, 2019.