Abstract: For the regenerative braking systems of electric vehicles， the corresponding regenerative braking control strategy might be designed according to the different braking intentions of the driver， so as to improve the safety， comfort and economy effectively during the vehicle braking processes. Accurate and fast recognition of drivers braking intentions was the basis of designing control strategy. An on-line recognition method of driver braking intentions was designed based on ABC-SVM algorithm  and implemented to identify the drivers braking intentions accurately and quickly for the electric vehicles with brake-by-wire. Firstly， the braking data were preprocessed， and the effective features were selected by NCA feature selection algorithm， and then the braking intention recognition model was established by the ABC-SVM algorithm， and on-line recognition was carried out finally. Offline verification and online test results show that the NCA algorithm may effectively filter out irrelevant features caused by signal noise. Compared with the fuzzy reasoning， back propagation（BP）， particle swarm optimization support vector machine（PSO-SVM）and genetic algorithm support vector machine （GA-SVM） recognition algorithms， the ABC-SVM algorithm may identify drivers braking intentions more accurately and quickly.

Key words: electric vehicle, braking intention, neighborhood component analysis（NCA）, artificial bee colony（ABC）, support vector machine（SVM）

摘要： 在电动汽车再生制动系统中，根据驾驶员不同的制动意图制定对应的再生制动控制策略可以有效地提高汽车的制动安全性、舒适性和经济性，而准确并快速识别驾驶员制动意图是制定控制策略的基础。以准确并快速识别驾驶员的制动意图为主要目标，以搭载线控制动系统的电动汽车为研究对象，设计并实现了一种基于人工蜂群支持向量机（ABC-SVM）的驾驶员制动意图在线识别方法。首先对制动数据进行预处理，用近邻成分分析（NCA）特征选择算法选取有效特征，再用ABC-SVM算法建立制动意图识别模型，最后进行在线识别。离线验证和在线试验结果表明，NCA算法能有效筛选掉信号噪声导致的不相关特征；相比于模糊推理、反向传播（BP）算法、粒子群优化支持向量机（PSO-SVM）和遗传算法支持向量机（GA-SVM）识别算法，ABC-SVM算法能够更加准确、快速地识别驾驶员的制动意图。

关键词: 电动汽车, 制动意图, 近邻成分分析, 人工蜂群, 支持向量机