基于多传感器信息融合的仿人机器人跌倒检测及控制

Falling Detection and Control of Humanoid Robots Based on Multi-Sensor Information Fusion

仿人机器人在行走过程中,需要及时检测到是否要摔倒,从而阻止摔倒或通过保护动作在摔倒过程中减少机器人损伤.文中基于姿态传感器(包括加速度传感器和陀螺仪)、力感应传感器(FSR)及步行阶段建立了多传感器信息融合模型,采用模糊逻辑决策方法建立了机器人跌倒检测的综合判定方法,并设计相应的控制器,该控制器可检测到仿人机器人运动时是否会摔倒,若无法避免摔倒,则采取相应的保护动作;若可避免摔倒,则通过控制髋关节阻止机器人摔倒.最后在SCUT-I仿人机器人上进行实验,结果表明,文中方法能及时检测到机器人在运动过程中发生跌倒的时刻,并及时通过稳定控制器阻止摔倒或通过倒地控制器产生保护动作,减少机器人的损伤.

When humanoid robots are walking, it is essential to detect in time whether the robots are going to fall, so as to prevent the robots from the falling or reduce the damage from the falling with the help of the corresponding protection actions. In this paper, a multi-sensor information fusion model in the stage of walking is constructed by using attitude sensors （including an acceleration sensor and a gyroscope） and force sensor resistors （FSR） . Then, a comprehensive judgment method of the falling detection of a humanoid robot is proposed by adopting fuzzy logic decision method, and the corresponding controllers are designed to detect whether the robot is going to fall. If the falling is unavoidable, the corresponding protection actions will be adopted. Otherwise, if the falling can be avoi-ded ,it will be prevented by controlling hip joints. The experimental results of SCUT-I humanoid robot show that the proposed method can detect the moment at which the robot is going to fall, and then prevent the falling by using the stability controller or reduce the damage to the robot by using the falling controller to bring about the corresponding protection actions.