Decomposition of spiking neural networks for hardware implementation of a mobile robot navigation system in an obstacle environment

Decomposition of spiking neural networks for hardware implementation of a mobile robot navigation system in an obstacle environment

Tim T. Isakov
Mathematician, Russian State Scientific Center for Robotics and Technical Cybernetics (RTC), 21, Tikhoretsky pr., Saint Petersburg, 194064, Russia, This email address is being protected from spambots. You need JavaScript enabled to view it., ORCID: 0000-0003-4437-5018

UDC identifier: 004.896

EDN: RXFZCA

Abstract. Recently, spiking neural networks have begun to be used for controlling mobile robots, including in energy-efficient hardware implementations. Specialized neuromorphic processors with a high level of parallelism are required for their hardware realization. One of the key challenges is transferring trained spiking neural network models to these processors, the efficiency of which directly influences the utilization of computational resources. This transfer involves distributing the model across processor cores, and it is suggested that dividing the model into several subnetworks, each solving a separate task, can simplify this process. This paper analyzes the relationship between network size and the quality of solving the dynamic obstacle avoidance task by a wheeled mobile robot under various complexity scenarios. A spiking neural network, trained using reinforcement learning algorithms, controls the drives of the wheeled robot by leveraging a priori data about the environment state from the simulator (including the speed and coordinates of both the obstacle and the robot, etc.). Based on the results obtained, we concluded that it is feasible to divide the network into smaller subnetworks, each effectively solving a simple task. Furthermore, a version of a neuromorphic control system using such a combination of networks is proposed.

Key words: spiking neural networks, reinforcement learning, dynamic obstacle avoidance, mobile robot, wheeled robot, neural network architecture

For citation: Isakov, T.T. (2025), "Decomposition of spiking neural networks for hardware implementation of a mobile robot navigation system in an obstacle environment", Robotics and Technical Cybernetics, vol. 13, no. 4, pp. 293-300, EDN: RXFZCA. (in Russian).

References

  1. Choi, J., Lee, G. and Lee, C. (2021), “Reinforcement learning-based dynamic obstacle avoidance and integration of path planning”, Intelligent Service Robotics, 14, pp. 663-677, DOI: 10.1007/s11370-021-00387-2
  2. Gao, X., Yan, L., Li, Z., Wang, G. and Chen, I.M. (2023), “Improved deep deterministic policy gradient for dynamic obstacle avoidance of mobile robot”, IEEE Transactions on Systems, Man, and Cybernetics: Systems, 53(6), pp. 3675-3682, DOI: 10.1109/TSMC.2022.3230666
  3. Tang, G., Kumar, N. and Michmizos, K.P. (2020), “Reinforcement co-learning of deep and spiking neural networks for energy-efficient mapless navigation with neuromorphic hardware”, In 2020 IEEE/RSJ International Conference on Intelligent Robots and Systems (IROS),pp. 6090-6097, DOI: 10.1109/IROS45743.2020.9340948
  4. Lillicrap, T.P. (2015), “Continuous control with deep reinforcement learning”, arXiv preprint, arXiv:1509.02971, DOI:10.48550/arXiv.1509.02971
  5. Wang, Y., Dong, B., Zhang, Y., Zhou, Y.et al. (2023), “Event-Enhanced Multi-Modal Spiking Neural Network for Dynamic Obstacle Avoidance”, In Proceedings of the 31st ACM International Conference on Multimedia, pp. 3138-3148, DOI: 10.1145/3581783.3612147
  6. Davies, M., Srinivasa, N., Lin, T.H., Chinya et al. (2018), “Loihi: A neuromorphic manycore processor with on-chip learning”, Ieee Micro, 38(1), pp. 82-99, DOI: 10.1109/MM.2018.112130359
  7. Akopyan, F., Sawada, J., Cassidy, A., Alvarez-Icaza, R., et al. (2015), “Truenorth: Design and tool flow of a 65 mw 1 million neuron programmable neurosynaptic chip”, IEEE transactions on computer-aided design of integrated circuits and systems, 34(10), pp.1537-1557, DOI: 10.1109/TCAD.2015.2474396
  8. MOTIV NT, “The first neuromorphic microprocessor in Russia”, available at: https://motivnt.ru/neurochip-altai/ (Accessed 22 November 2024). (in Russian).
  9. Schulman, J., Wolski, F., Dhariwal, P., Radford, A. and Klimov, O. (2017), “Proximal policy optimization algorithms”, arXiv preprint, arXiv:1707.06347, DOI: 10.48550/arXiv.1707.06347
  10. Rojas, M., Hermosilla, G., Yunge, D. and Farias, G. (2022), “An Easy to Use Deep Reinforcement Learning Library for AI Mobile Robots in Isaac Sim”, Applied Sciences, 12(17), p. 8429, DOI: 10.3390/app12178429

Received 17.07.2025
Revised 20.07.2025
Accepted 04.08.2025