基于改进头姿估计方法的机器人轮椅交互控制

针对现有迭代最近点（ICP）头姿估计算法存在迭代次数偏多且易陷于局部最优、而随机森林（RF）头姿估计算法准确性和稳定性不高的问题，提出一种新的头姿估计改进方法，并基于该改进方法构建机器人轮椅实时交互控制接口．首先，分析现有迭代最近点头姿算法与随机森林头姿算法在准确性、实时性及稳定性方面存在的问题，并提出一种新的基于随机森林与迭代最近点算法融合的头姿估计改进方法；其次，为实现头姿估计到机器人轮椅交互控制的无缝连接，建立基于传统机器人轮椅操纵杆的头部姿态运动空间映射；最后，在基于标准头姿数据库分析改进头姿估计方法性能的基础上，构建机器人轮椅实验平台并规划运动轨迹，以进一步验证基于改进头姿估计方法的人机交互接口在机器人轮椅实时控制方面的有效性．实验结果表明，改进后的头姿估计方法较传统迭代最近点算法减少了迭代次数且避免了陷于局部最优，在仅增加少量运算时间的基础上，其准确性和稳定性都优于传统随机森林算法；同时，基于改进头姿估计方法的人机交互接口亦能实时平稳地控制机器人轮椅沿既定的轨迹运动．

Interactive Control of Robotic Wheelchair Based on an Improved Head Pose Estimation Method

As for the problem that the current iterative closet point (ICP) based head pose estimation method has more iterative steps and easily falls into local optimum while the random forest (RF) based head pose estimation method is of low accuracy and inferior stability, an improved head pose estimation method is proposed, and a real-time interactive control interface of robotic wheelchair based on the improved head pose estimation method is designed. Firstly, an improved head pose estimation method using ICP and RF algorithms is proposed based on the analyses on accuracy, real-time performance and stability issues existed in the current ICP and RF based head pose estimation algorithms. Then, the traditional robotic wheelchair joystick based head pose motion space mapping is built to achieve the seamless connections from the head pose estimation to the interactive control of robotic wheelchair. Finally, based on the performances analyses of the improved head pose estimation method using the standard head pose database, the robotic wheelchair platform is set up, and the movement trajectories are planned to verify the effectiveness of the proposed human-robot interactive interface using the improved head pose estimation method on the real-time control of robotic wheelchair. The experimental results demonstrate that the improved head pose estimation method has less iterative steps and can avoid falling into local optimum compared with the traditional ICP method, and the accuracy and stability of the proposed algorithm are better than the traditional RF method by increasing only a small amount of computation time. Moreover, the human-robot interactive interface using the improved head pose estimation method can control the robotic wheelchair following the predefined trajectories in a real-time and smooth manner.