基于场景理解与改进型BUG算法的移动机器人避障

针对现有移动机器人在视觉避障上存在的局限,将深度学习算法和路径规划技术相结合,提出了一种基于深层卷积神经网络和改进Bug算法的机器人避障方法。该方法采用多任务深度卷积神经网络提取道路图像特征,实现图像分类和语义分割任务；其次,基于语义分割结果构建栅格地图,并将图像分类结果与改进的Bug算法相结合,搜索出最优避障路径；同时,为降低冗余计算,设计了特征对比结构来对避免对重复计算的特征信息,保障机器人在实际应用中实时性。通过实验结果表明,所提方法有效的平衡了多视觉任务的精度与效率,并能准确规划出安全的避障路径,辅助机器人完成导航避障。

Mobile Robot Obstacle Avoidance Based on Deep Learning and Improved Bug Algorithm

Aiming at the limitations of existing mobile robots in visual obstacle avoidance, a robot obstacle avoidance method based on deep convolutional neural network and Bug algorithm is proposed by combining deep learning algorithm and path planning technology. In this method, multi task deep convolution neural network is used to extract road image features to realize image classification and semantic segmentation; Secondly, the grid map is constructed based on the semantic segmentation results, and the image classification results are combined with the improved bug algorithm to search the optimal obstacle avoidance path; At the same time, in order to reduce redundant calculation, a feature comparison structure is designed to avoid the feature information of repeated calculation and ensure the real-time performance of the robot in practical application. The experimental results show that the proposed method effectively balances the accuracy and efficiency of multi-vision tasks, and can accurately plan a safe obstacle avoidance path to assist the robot to complete navigation and obstacle avoidance.