摘要
针对已有传统避障算法的各种缺陷,提出一种矢量场直方图法(Vector FieldHistogram, VFH)结合神经网络(Neural Network)算法的新型混合实时避障策略.该策略利用矢量场直方图法实现基础避障,得到大量数据集.利用样本集训练神经网络可以得到一个神经网络预测器.从而实现同时具有预测和控制运动矢量的效果.为了验证该策略的有效性,针对不同障碍环境下进行仿真,并与单一避障效果进行了对比.仿真实验验证了这种实时避障策略的可行性和优越性.
A new hybrid real-time obstacle avoidance strategy based on VFH (Vector Field Histogram) and neural network algorithm is proposed according to traditional obstacle avoidance algorithm. In this strategy the vector field histogram method is used to achieve obstacle avoidance firstly, while obtaining a large number of data sets. And a neural network predictor can be obtained by training the data sets. Thereby by this algorithm it can achieve the effect of simultaneously predicting and controlling the motion vector. In order to verify the effectiveness of the strategy, simulation is carried out for different obstacles, and comparing with the single obstacle avoidance. The simulation experiment verifies the feasibility and superiority of this real-time obstacle avoidance strategy.
作者
张倩倩
余道洋
李民强
ZHANG Qian-qian;YU Dao-yang;LI Min-qiang(Institute of Intelligent Machines, Chinese Academy of Sciences, Hefei 230031, China;Department of Automation, University of Science and Technology of China, Hefei 230026, China)
出处
《控制工程》
CSCD
北大核心
2019年第7期1328-1334,共7页
Control Engineering of China
基金
功能化的纳米间隙电极及其在低浓度乙烯快速检测中的应用(31571567)
关键词
移动机器人
矢量场直方图算法
神经网络
混合策略避障控制
Mobile robot
Vector field histogramm
neural network
hybrid strategy of obstacle avoidance control
