Research On Algorithms For Biped Motion

ABSTRACT:The algorithms of Biped walking which fit the physical law of the real world,have been a hot spot and difficult point of graphics and robotics for more than20years.It has been applied widely in almost all the aspects of our society,such as military affairs,technology,home entertainment,dissemination of information,making the research value very tremendous.Because the widely used PD control cannot eliminate the static error, which belongs to an algorithm of sacrificing accuracy for real-time performance, but the rapid development of the hardware facilities make the advantage less obvious, the PID algorithm is introduced to eliminate static error basically, in order to improve the accuracy of the overall system.And because the biped humanoid robot model itself belongs to the high dimension nonlinear system, while artificial neural network algorithm has computational advantages for handling nonlinear data, its self-learning characteristics can also strengthen its natural and further improve the precision of calculation, the use of artificial neural network with BP algorithm as the core can improve the accuracy of calculation.After introducing PID artificial neural network algorithm based on BP algorithm as the core into the basic control part of the classic SIMBICON, the realization of a complete3D character motion system is more natural for basic walking speed. Walking characteristics of the improved system especially in the conventional amplitude and of small amplitude motion is more close to the human,the naturality and the accuracy is better.