动量-自适应BP算法在机器人碰撞检测仿真系统中的应用

根据机器人运动连续性原理,通过对误差脉冲数的统计分析,我们基于人工神经网络算法,实现了机器人碰撞检测仿真系统 根据从机器人运行时采集的数据对神经网络进行训练和仿真,在实际应用中取得了预期的效果 本文讨论了动量-自适应学习率BP算法,说明了通过误差脉冲数进行碰撞检测的原理,比较了它与传统方法的区别,并且根据神经网络训练和仿真结果对动量-自适应学习率BP算法和标准BP算法进行了比较.

Adaptive-Learning-Rate with Momentum BP Algorithm and Its Application in the Simulating System of Robot Collision Detection

According to the running-continuous principle of the robot and the statistics and analysis of the error count, we developed a robot collision detection simulating system based on artificial neural net. We trained the neural net with momentum-adaptive learning rate BP algorithm and simulated the collected data. The results show that the neural net can meet our need for collision detection. We discussed adaptive-learning-rate with momentum BP algorithm in this paper, and compared it with standard BP algorithm according to the results of training and simulation of the neural net. We also explained the collision detection method using the error count, and compared it with traditional method.