The use of llm models on single-board computers for the implementation of autonomous uav flight
- Authors: Anisimov R.O.1, Dvornikov A.D.1, Kulagin K.A.1, Titova S.A.2, Petrov K.V.2
-
Affiliations:
- V.A. Trapeznikov Institute of Control Sciences of RAS
- Moscow Polytechnic University
- Issue: No 117 (2025)
- Pages: 246-264
- Section: Information technologies in control
- URL: https://bakhtiniada.ru/1819-2440/article/view/360566
- DOI: https://doi.org/10.25728/ubs.2025.117.12
- ID: 360566
Cite item
Full Text
Abstract
This paper examines the use of large language models (LLM) to control unmanned aerial vehicles (UAVs) using natural language commands. The research is aimed at solving a key problem – the discrepancy between the high computing requirements of LLM and the limited resources of on-board computers. The main focus is on optimizing LLM for operation on energy-efficient single-board computers with neuroprocessors, such as OrangePi 5B based on Rockchip RK3588S. The paper presents the results of testing Qwen2.5-Coder quantized models, demonstrating the preservation of code generation quality at processing speeds of up to 17.8 tokens/s. A specialized test (benchmark) was developed to evaluate the effectiveness of LLM integration into UAV autonomous control architecture and the correctness of code generation, including 125 scenarios. The results confirm the feasibility of LLM in autonomous drone control systems, although they reveal typical errors associated with sensor data processing and coordinate systems. The study proposes a promising direction for the development of intelligent UAV control systems with a natural language interface (NLP). The study included both a scientific approach (development of a specialized test) and a technological innovation (performance analysis on single-board computers) aimed at integrating LLM into UAV autonomous control architecture.
About the authors
Rodion Olegovich Anisimov
V.A. Trapeznikov Institute of Control Sciences of RAS
Email: rodion_anisimov@mail.ru
Moscow
Alexey Dmitrievich Dvornikov
V.A. Trapeznikov Institute of Control Sciences of RAS
Email: applskyp@gmail.com
Moscow
Konstantin Aleksandrovich Kulagin
V.A. Trapeznikov Institute of Control Sciences of RAS
Email: kka8686@mail.ru
Moscow
Sofya Alekseevna Titova
Moscow Polytechnic University
Email: titovas63059@gmail.com
Moscow
Konstantin Vladimirovich Petrov
Moscow Polytechnic University
Email: r.92rab@gmail.com
Moscow
References
1. ARIZA J.A., BAEZ H. Understanding the role of single-board computers in engineering and computer science edu-cation: a systematic literature review // Computer Applica-tions in Engineering Education. – 2022. – Vol. 30(1). – P. 304–329. – doi: 10.1002/cae.22439.2. AUSTIN J., OZI A., DRAZNER S. Program synthesis with large language models // Journal of Artificial Intelligence Research. – 2021. – Vol. 71. – P. 355–420.3. BROWN T.B. et al. Language models are few-shot learners // arXiv:2005.14165. – 2020. – 75 p.4. CHEN K., WILLIAMS D., TAYLOR M. SparseGPT: prun-ing large language models for embedded applications // Journal of Artificial Intelligence Research. – 2024. – Vol. 55. – P. 213–229.5. CHEN M., TWIGHT B., LEWIS M. Evaluating large lan-guage models trained on code // arXiv:2107.03374. – 2021.6. DETTMERS T., ZETTLEMOYER L., LEWIS M. LLM.int8(): 8-bit matrix multiplication for transformers at scale // arXiv:2208.07339. – 2022.7. JOHNSON T., LEE M., CARTER P. Autonomous aerial ex-ploration using compact language models. MIT computer science and artificial intelligence laboratory // Technical Report MIT-CSAIL-TR-2024-03. – 2024.8. KIM S., PAK D., LEE J. Energy consumption analysis of LLM-powered UAVs // Journal of Unmanned Aerial Systems. – 2024. – Vol. 12, No. 3. – P. 178–192.9. KIM Y., PARK J., SHIN J. PQK: practical quantization and pruning for on-device vision transformers // arXiv:2106.14681. – 2021.10. LEE D., SMITH R., PARKER A. MiniLLM: extreme quanti-zation for large language models on resource-constrained platforms // Proc. of the Conf. on Machine Learning and Systems. – 2024. – P. 1243–1255.11. LIU D., LI H., SHEN Y., et al. KeepEdge: lightweight real-time detection for UAV applications // IEEE Trans. on Indus-trial Informatics. – 2023. – Vol. 19(4). – P. 2873–2885.12. MULLER A., SCHMIDT K., WEBER R. Emergency re-sponse drone system with on-board decision making // ETH Zurich Robotics and Perception Group Research Reports. – 2023.13. RAMIREZ M., AL-HADITI F., NEIM S. DroneGPT: natural language control for unmanned aerial vehicles // Stanford AI Lab Technical Report, SAIL-TR-2023-04. – 2023.14. SU J. et al. RoFormer: enhanced transformer with rotary position embedding // arXiv:2104.09864. – 2021.15. TING L., JOHNSON M., CHEN K. Performance evaluation of edge computing platforms for AI applications in UAVs // IEEE Trans. on Robotics and Automation. – 2023. – Vol. 38, No. 2. – P. 112–127.16. VAN X., BROWN T., ANDERSON J. DistilLLM: knowledge distillation methods for efficient deployment of language models in UAVs // Int. Conf. on Robotics and Automation (ICRA). – 2023. – P. 775–782.17. VASWANI A. et al. Attention is all you need // 31st Conf. on Neural Information Processing Systems (NIPS–2017), Long Beach, CA, USA. – arXiv:1706.03762v7. – 2 Aug 2023. – P. 1–15.18. ZHANG Y., LI W., MARTINEZ K. Hybrid LLM-traditional architectures for autonomous UAV navigation // Robotics and Autonomous Systems Journal. – 2024. – Vol. 152. – P. 45–62.
Supplementary files
