【摘要】 针对复杂步态下柔性外骨骼虚拟惯性测量组件(VIMU)性能稳定性下降的问题,研究了一种基于支持向量机(SVM)步态分类的自主定位优化方法。采用SVM算法模型对柔性外骨骼的多种常规步态类型进行识别,根据不同的步态类型构建不同的卷积-长短期记忆(VGG-LSTM)混合神经网络模型,并通过判断实际惯性测量组件(IMU)的故障,利用VIMU构成具备系统重构能力的强鲁棒性自主定位方法。研究结果表明,复杂步态下基于SVM的步态分类方法可在保证VIMU精度的同时,降低VGG-LSTM神经网络模型的复杂性;机器人肢节末端IMU在常规步态下出现故障时,系统重构后的自主定位性能与无故障情况下基本保持一致,重构导航系统的定位误差在行进距离的2.5%以内。更多还原

【Abstract】 Aiming at the problem that the stability of virtual inertial measurement unit(VIMU) of the flexible exoskeleton decreases in complex gait, an autonomous positioning optimization method based on support vector machine(SVM) gait classification was studied. The SVM algorithm model was used to identify multiple conventional gait types of the flexible exoskeleton, and different convolution-long-short-term memory(VGG-LSTM) mixed neural network models were constructed according to the gait types. By judging the failure of the actual inertial measurement unit(IMU), the VIMU was used to form a robust and autonomous positioning method with system reconfiguration capabilities. The research result shows that, the gait classification method based on SVM in complex gaits can reduce the complexity of the VGG-LSTM neural network model while ensuring the accuracy of VIMU. When the IMU at the end of the robot limb fails in conventional gaits, the autonomous positioning performance of the system after reconstruction is basically the same as that without failures. The positioning error of the reconstructed navigation system is less than 2.5% of the travel distance.更多还原