Publisher

IAESONLINEIAESONLINE

Journal Indonesian Journal of Electrical Engineering and Informatics (IJEEI)Indonesian Journal of Electrical Engineering and Informatics (IJEEI) Abstract

Partial observability and sensor limitations are challenging for the navigation of autonomous Unmanned Aerial Vehicles (UAVs). Deep Reinforcement Learning (DRL) algorithms have emerged as potential tools in advancing this field. However, their effectiveness degrades in challenging environments, particularly in the presence of dynamic obstacles. Recent research trends emphasize the need for new DRL variants that guarantee robustness, real-time adaptability, and improved generalization under uncertainty. This paper proposes a lightweight DRL architecture that combines Proximal Policy Optimization (PPO) with a Gated Recurrent Unit (GRU), extended with a temporal LiDAR differencing feature called Delta-LiDAR. The difference between consecutive LiDAR scans is computed to provide the velocity and directional cues without the computational burden of Long Short-Term Memory (LSTM) networks. We evaluate three models, PPO-LSTM, PPO-GRU, and Delta-LiDAR augmented PPO-GRU in a 3D simulated UAV navigation environment characterized by noise, clutter, and dynamic obstacles. We considered several metrics, including success rate, collision frequency, trajectory smoothness, and computational efficiency, to determine the effectiveness of each architecture. The experimental results demonstrate that Delta-LiDAR improves GRU-based temporal reasoning. The deployment complexity is reduced compared with the LSTM-based architecture, which makes it ideal for real-time UAV operation in partially observable environments.

Conclusion

This study presented a novel approach to enhance UAV navigation in partially observable and dynamic environments by integrating Delta-LiDAR with a GRU-based PPO framework.The proposed method effectively captures motion cues from LiDAR scans, improving temporal awareness without increasing computational complexity.Experimental results demonstrate that the PPO-GRU with Delta-LiDAR outperforms both PPO-LSTM and standard PPO-GRU in terms of success rate, collision avoidance, and trajectory efficiency.The findings validate the potential of this approach for real-time UAV navigation in challenging environments.

Future Research

Penelitian lebih lanjut dapat dilakukan untuk mengeksplorasi integrasi model hierarkis atau berbasis Transformer dengan Delta encoding untuk perencanaan misi UAV yang lebih kompleks dalam lingkungan yang dinamis dan tidak pasti. Selain itu, penelitian dapat difokuskan pada peningkatan kemampuan persepsi dan perencanaan dengan menggabungkan input sensor tambahan seperti data visual, radar, atau kamera berbasis event. Terakhir, untuk mengatasi keterbatasan dalam respons terhadap perubahan lingkungan yang sangat cepat, penelitian dapat menyelidiki optimasi kecepatan pengambilan sampel sensor dan kecepatan pemrosesan onboard, serta pengembangan algoritma yang lebih adaptif terhadap perubahan dinamis. Pengembangan ini akan memungkinkan UAV untuk beroperasi secara lebih otonom dan efisien dalam berbagai skenario dunia nyata, termasuk pencarian dan penyelamatan, inspeksi infrastruktur, dan pengawasan lingkungan. Penerapan teknik pembelajaran transfer juga dapat dieksplorasi untuk mempercepat proses pembelajaran dan meningkatkan generalisasi model ke lingkungan yang berbeda. Integrasi dengan sistem manajemen lalu lintas udara (UTM) juga dapat menjadi arah penelitian yang menjanjikan untuk memastikan operasi UAV yang aman dan terkoordinasi.

References

1. Enhancing GRU-Based DRL with Delta-LiDAR for Robust UAV Navigation in Partially Observable Dynamic Environments... doi.org/10.52549/ijeei.v13i3.7067Enhancing GRU Based DRL with Delta LiDAR for Robust UAV Navigation in Partially Observable Dynamic Environments doi 10 52549 ijeei v13i3 7067

Metadata

File size	1.5 MB
Pages	16
DMCA	Report