基于脉冲神经网络的智能小车的自动避障系统研究
Research on Automatic Obstacle Avoidance System of Smart Vehicle Based on Spike Neural Network
摘要: 为了提高智能小车自动避障控制的精度，本文提出了一种基于脉冲神经网络(SNN)的避障控制模型。该神经网络采用泊松编码方式，使网络模型不仅具有更明确的生物意义，还具有强大的计算能力。本文针对全连接网络结构导致计算量大的问题提出了一种基于STDP规则的稀疏概率连接网络结构，通过实验验证，该结构有效地减少了网络的训练时间，同时还提高了算法的效率和准确率。采用多个超声波传感器探测障碍物的方位，信息经过脉冲神经网络处理后，实现了智能小车对障碍物的安全避障控制。考虑小车与障碍物和目标点的距离、角度以及小车的行驶速度等多种路况，该模型采用灰色关联分析法选取相似路况进行模拟验证。本文分别对基于SNN、支持向量机(SVM)和BP人工神经网络(BP-ANN)的控制模型进行测试和评估。实验结果表明：SNN避障控制模型相比于其他模型有更高的控制精度和适用性，为智能车辆的开发避障控制算法提供了一定的理论依据和应用价值。
Abstract: An obstacle avoidance control model based on Spiking Neural Network (SNN) was proposed in this paper to improve the accuracy of the automatic obstacle avoidance control of the smart vehicle. This neural network uses Poisson coding, which not only has a clearer biological meaning, but also has powerful computing capabilities. This paper proposes a sparse probability connection network structure based on STDP rules for the problem of a large amount of calculation caused by the fully connected network structure. Through experimental verification, this structure effectively reduces the training time of the network and also improves the efficiency and accuracy of the algorithm. Multiple ultrasonic sensors are used to detect the orientation of obstacles. After the information is processed by the pulse neural network, the smart car realizes the safe obstacle avoidance control of the obstacles. Considering the road conditions such as the distance and angle between the trolley and the obstacle and the target point, as well as the traveling speed of the trolley, the model adopts the gray correlation analysis method to select similar road conditions for simulation verification. This paper tests and evaluates control models based on SNN, support vector machine (SVM), fuzzy neural network and BP artificial neural network (BP-ANN). The experimental results show that the SNN obstacle avoidance control model has higher control accuracy and applicability than other models, which provides a certain theoretical basis and application value for the development of obstacle avoidance control algorithms for intelligent vehicles.
文章引用：陈维, 陈靖宇. 基于脉冲神经网络的智能小车的自动避障系统研究[J]. 计算机科学与应用, 2021, 11(5): 1445-1456. https://doi.org/10.12677/CSA.2021.115148

参考文献

[1] 朱大奇, 孙兵, 李利. 基于生物启发模型的AUV三维自主路径规划与安全避障算法[J]. 控制与决策, 2015, 30(5): 798-806.
[2] 张鹏彬, 曾钰培. 改进人工势场法的移动机器人路径规划[J]. 机器人技术与应用, 2018(3): 27-29.
[3] Shu, L., Yang, J., Li, B., Deng, Z. and Dapino, M.J. (2020) Impact Force Sensing with Magnetostrictive Fe-Ga Alloys. Mechanical Systems and Signal Processing, 139, Article ID: 106418. [Google Scholar] [CrossRef] 
[4] 汤荣志, 段会川, 孙海涛. SVM训练数据归一化研究[J]. 山东师范大学学报(自然科学版), 2016, 31(4): 60-65.
[5] 杨亚兰, 徐耀良, 钟绍山, 谢江媛. 风电场短期风速预测的MRASVM模型[J]. 电力系统及其自动化学报, 2014, 26(5): 44-49.
[6] 丁明, 王伟胜, 王秀丽, 宋云亭, 陈得治, 孙鸣. 大规模光伏发电对电力系统影响综述[J]. 中国电机工程学报, 2014, 34(1): 1-14.
[7] 殷桂梁, 李相男, 郭磊, 李晓亮. 混合储能系统在风光互补微电网中的应用[J]. 电力系统及其自动化学报, 2015, 27(1): 49-53, 59.
[8] 杨晓萍, 王宝, 兰航, 武小暄. 风电场短期功率预测[J]. 电力系统及其自动化学报, 2015, 27(9): 85-90.
[9] Fardsoleimani, L., Khanmohammadi Khorrami, M., Zandbaaf, S., Rezanejade Bardajee, G. and Ghasemi, K. (2020) Determination of Micropore Volumes of Nano ZSM-5 Zeolite Samples by Diffuse Reflectance Infrared Fourier Transform (DRIFT) Spectroscopy Using Back-Propagation Artificial Neural Network (BP-ANN) and Non-Negative Matrix Factoriza-tion-Alternating Least Squares (NMF-ALS) as Chemometric Approaches. Infrared Physics and Technology, 111, Article ID: 103543. [Google Scholar] [CrossRef] 
[10] 盛贤君, 姜涛, 王杰, 刘宁. 基于BP神经网络的速度前馈PID控制器设计[J]. 人工智能, 2015, 35(z2): 134-137.
[11] Maass, W. (1997) Networks of Spiking Neurons: The Third Gen-eration of Neural Network Models. Neural Networks, 10, 1659-1671. [Google Scholar] [CrossRef] 
[12] Pavlidis, N.G., Tasoulis, O.K., Plagianakos, V.P., et al. (2005) Spik-ing Neural Network Training Using Evolutionary Algorithms. 2005 IEEE International Joint Conference on Neural Networks, Montreal, 31 July-4 August 2005, 2190-2194.
[13] Wei, K.X., Lauer, I., Srinivasan, S., Sundaresan, N., McClure, D.T., Toyli, D., et al. (2020) Verifying Multipartite Entangled Greenberger-Horne-Zeilinger Statesvia Multiple Quantum Coherences. Physical Review A, 101, Article ID: 032343. [Google Scholar] [CrossRef] 
[14] Feng, J. and Brown, D. (2000) In-tegrate-and-Fire Models with Nonlinear Leakage. Bulletin of Mathematical Biology, 62, 467-481. [Google Scholar] [CrossRef] [PubMed]
[15] Hunsberger, E. and Eliasmith, C. (2015) Spiking Deep Networks with LIF Neurons. arXiv: 1510.08829.
[16] 程龙, 刘洋. 脉冲神经网络: 模型、学习算法与应用[J]. 控制与决策, 2018, 33(5): 923-937.
[17] Heeger, D. (2000) Poisson Model of Spike Generation. Handout, University of Standford, 5, 76.
[18] Bellis, S., Razeeb, K.M., Saha, C., Delaney, K., O’Mathuna, C., Pounds-Cornish, A. and Nikiforidis, G. (2004) FPGA Implementation of Spiking Neural Networks—An Initial Step towards Building Tangible Collaborative Autonomous Agents. 2004 IEEE Interna-tional Conference on Field-Programmable Technology (IEEE Cat. No. 04EX921), Brisbane, 6-8 December 2004, 449-452.
[19] 蔺想红, 王向文, 张宁, 马慧芳. 脉冲神经网络的监督学习算法研究综述[J]. 电子学报, 2015, 43(3): 577-586.
[20] Querlioz, D., Bichler, O., Dollfus, P. and Gamrat, C. (2013) Immunity to Device Variations in a Spiking Neural Network with Memristive Nanodevices. IEEE Transactions on Nanotechnology, 12, 288-295. [Google Scholar] [CrossRef] 
[21] Izhikevich, E.M. (2007) Solving the Distal Reward Problem through Linkage of STDP and Dopamine Signaling. Cerebral Cortex, 17, 2443-2452. [Google Scholar] [CrossRef] [PubMed]
[22] 刘硕, 师五喜. 智能小车编队避障控制系统设计[J]. 自动化与仪表, 2019, 34(1): 28-32.
[23] 王守相, 张娜. 基于灰色神经网络组合模型的光伏短期出力预测[J]. 电力系统自动化, 2012, 36(19): 37-41.
[24] 丁世飞, 齐丙娟, 谭红艳. 支持向量机理论与算法研究综述[J]. 电子科技大学学报, 2011, 40(1): 2-10.
[25] 梁惠施, 程林, 苏剑. 微网的成本效益分析[J]. 中国电机工程学报, 2011, 31(z1): 38-44.