基于改进机器学习的图书馆机器人自主避障控制研究

Li Jing1, Luo Zheng 2,Yan ZhenPing 3 (1. Image and Text Information Center of Xi"an Eurasian University, Xi’an,710065,China;

摘要:

为控制图书馆机器人在行进过程中自动躲避障碍,达到理想工作效果,提出基于改进机器学习的图书馆机器人自主避障控制方法。利用安装于图书馆机器人机身的各组超声波测距传感器,采集图书馆机器人与目标障碍物距离信息,并将各组传感器采集到最小距离信息组合形成感知环境特征向量,当成卷积神经网络输入,经卷积、池化等操作,输出图书馆机器人对当前环境感知结果,同时采用粒子群算法改进卷积神经网络参数,而后把最优环境识别结果与该结果下图书馆机器人距障碍物间距、图书馆机器人实时速度,当成模糊PID控制模型有效输入,经输入输出变量模糊化、模糊推理以及输出变量解模糊等操作后,实现图书馆机器人自主避障无冲突运行。实验结果表明：采用该方法进行避障控制,自主避障控制效果较好,在远距离即可作出反应,避障行驶距离短,高速运行时反应更快；在复杂静态环境中,机器人能稳定、平滑地避开多个障碍物,到达终点；且该方法识别分类结果与实际感知环境类型一致,为图书馆机器人自主避障提供可靠保障。

Abstract:

Propose a library robot autonomous obstacle avoidance control method based on improved machine learning, which controls the library robot to automatically avoid obstacles during movement and achieve ideal working results. By using various sets of ultrasonic distance sensors installed on the body of the library robot, the distance information between the library robot and the target obstacle is collected. The minimum distance information collected by each set of sensors is combined to form a perception environment feature vector, which is used as input for the convolutional neural network. After convolution, pooling and other operations, the perception results of the library robot on the current environment are output. At the same time, particle swarm optimization algorithm is used to improve the convolutional neural network parameters. Then, the optimal environment recognition result is combined with the distance between the library robot and the obstacle and the real-time speed of the library robot under this result as effective inputs for the fuzzy PID control model. After input and output variable fuzzification, fuzzy reasoning, and output variable deblurring operations, the library robot achieves autonomous obstacle avoidance and conflict free operation. The experimental results show that using this method for obstacle avoidance control, the autonomous obstacle avoidance control effect is good, and the response can be made at a long distance, the obstacle avoidance driving distance is short, and the response is faster when running at high speed; In a complex static environment, the robot can avoid multiple obstacles stably and smoothly and reach the finish line. Moreover, the recognition and classification results of this method are consistent with the actual perceived environment types, which provides a reliable guarantee for library robots to avoid obstacles autonomously.