应用于视觉环境探索的移动机器人反应式控制器设计

在移动机器人的视觉环境探索研发中,在视觉定位不精确的条件下,如何保证移动机器人可靠地运动,是目前移动机器人视觉环境探索与视觉导航中有待解决的问题。提出一种基于单帧图像语义分割的运动控制算法,利用深度卷积神经网络高鲁棒性的语义分割结果,在图像像素空间进行运动目标点规划。仅以当前观测图像为基础,以最大化探索可能运动方向为规划代价函数,在脱离全局视觉定位结果的情况下,实现移动机器人在视觉环境探索中的可靠运动控制以及避障。实验和仿真结果表明:所提出方法可满足环境探索问题中对环境覆盖以及避障的运动控制需要。

Design of Response Controller for Mobile Robot Applied to Visual Environment Exploration

In the exploration and development of the visual environment of a mobile robot,how to ensure the reliable movement of a mobile robot under the condition of imprecise visual positioning is a problem to be solved in the exploration and navigation of the mobile robot’s visual environment.A motion control algorithm based on a single frame image semantic segmentation was proposed.Based on the robust semantic segmentation results of deep convolutional neural network,the moving target point was planned in the pixel space of the image.On the basis of the current observed images and taking the maximization of possible motion directions as the planning cost function,the reliable motion control and obstacle avoidance of mobile robots in visual environment exploration are realized without the results of global visual positioning.The results show that the proposed method can meet the requirements of environmental coverage and obstacle avoidance in environmental exploration.