论文部分内容阅读
被动行走机器人由于结构简单、能量利用率高而倍受青睐,但其很容易跌倒,因此准确把握最终步态与吸引区域成了关键.由于面对非光滑系统,大规模数值计算很难避免,为此本文先提出基于CPU+GPU异构平台的Poincare映射算法.该算法可发挥最新平台计算潜力,比传统CPU上算法快上百倍.得益于此,本文针对双足被动行走的最基本模型,大规模地选取样点进行计算,不仅清晰地得出吸引区域的形状轮廓和细节特征,揭示了其内在分形结构,还得到系统吸引集和吸引区域随倾角k的变化关系,发现了新的稳定三周期步态和倍周期分岔混沌现象,并研究了吸引区域.
Passive walking robots are popular because of their simple structure and high energy efficiency, but they are easy to fall over. Therefore, it is crucial to accurately grasp the final gait and the attraction area.Because of the non-smooth system, large-scale numerical calculation is difficult to avoid, Therefore, this paper first proposes a Poincare mapping algorithm based on CPU + GPU Heterogeneous platform.This algorithm can take advantage of the latest platform computing potential, which is 100 times faster than the traditional CPU algorithm.Because of this, this paper focuses on the most basic model of passive biped walking , Large-scale selection of samples for calculation, not only clearly draw the shape of the suction area contour and the details of the features revealed its internal fractal structure, but also get the system to attract the region and the region of variation with the inclination k, found a new Stable three-cycle gait and double-cycle bifurcation chaos phenomenon, and to study the attractive area.