基于遗传算法的动态姿势预测模型的研究

Dynamic posture prediction model based on genetic algorithm

分析了人体姿势控制的一般研究方法,通过实验研究了姿势选择的影响因子,以及影响因子重要度的比较情况。提出了基于价值函数优化方法,用以模拟现实中人的行为。定义了相应的价值函数,并将这些价值函数用到姿势预测算法中。采用遗传算法来处理姿势预测问题,将舒适度、重力势能和关节活动范围作为约束条件,使姿势预测更加逼真。试验表明,该方法能解决数字化人机工程中的姿势仿真问题。

Traditional research methods for postural control were analyzed. Influencing factors on posture selection through experiments were studied, and importance of these factors were compared. Then, cost functions optimization was put forward to simulate the man＇s behavior in reality. Cost functions for posture prediction were defined and applied to algorithm. Genetic algorithm was adopted to deal with posture prediction. Comfort, potential energy and joint ranges of motion were set as constraints, as a result the posture prediction was more realistic. Experiment proved that the model could solve the reachable problem of space point in computer aided ergonomics, and predict reasonable posture.