A Self-Driving Car Traffic Risk Prediction Model Based on ANFIS

为降低交通事故风险并提升自动驾驶车辆的通行安全性和效率，研究了一种基于自适应神经模糊推理系统(ANFIS)的自动驾驶交通风险预判方法。首先，以车辆速度和加速度作为量化指标，构建交通风险预判模型的输入参数。其次，利用MATLAB工具箱对数据进行模糊化处理，通过模糊规则确定输入隶属度，计算隶属度的可信度及规则的综合适应度，并结合人工神经网络的训练能力，优化模型的决策质量和风险管理能力。最后，通过仿真验证，结果表明该模型的交通风险预测误差约为13.77%，误差较小，能够有效实现自动驾驶汽车的交通风险预判，为后续优化自动驾驶汽车性能、确保其行驶安全性和可靠性提供了有力支持。To reduce the risk of traffic accidents and enhance the safety and efficiency of autonomous vehicles, a method for traffic risk prediction in autonomous driving based on the Adaptive Neuro-Fuzzy Inference System (ANFIS) has been studied. Firstly, vehicle speed and acceleration are used as quantitative indicators to construct the input parameters of the traffic risk prediction model. Secondly, the data is fuzzified using the MATLAB toolbox. The membership degrees of the inputs are determined through fuzzy rules, and the credibility of the membership degrees and the comprehensive adaptability of the rules are calculated. Combining this with the training capability of artificial neural networks, the decision quality and risk management capability of the model are optimized. Finally, the model is verified through simulation, and the results show that the traffic risk prediction error of the model is approximately 13.77%, which is relatively low. This demonstrates the model’s effectiveness in predicting traffic risks for autonomous vehicles and provides strong support for further optimizing the performance of autonomous vehicles and ensuring their driving safety and reliability.