应用BP神经网络对冗余度二重八面体变几何桁架机器人的位姿正解进行训练学习 ,解决了二重八面体变几何桁架机器人的位姿反解问题。并利用网络模型求得运动学中的一、二阶影响系数矩阵 ,将关节极限优化指标引入机器人的运动学规划中 ,提出了一种轨迹规划方法。并对双二重八面体变几何桁架机器人进行了仿真计算 The BP neural network is used to train and learn the pose and attitude of redundancy dual octahedral variable geometry truss robots, which solves the problem of inverse pose and pose of double octahedral variable geometry truss robots. The network model was used to calculate the first and second order influence coefficient matrices, and the joint limit optimization index was introduced into the robot kinematics planning. A trajectory planning method was proposed. The double double octahedral variable geometry truss robot is simulated