基于行驶空间的地面无人平台越野自主导航研究

项目名称： 基于行驶空间的地面无人平台越野自主导航研究

项目编号： No.61473042

项目类型： 面上项目

立项/批准年度： 2015

项目学科： 自动化技术、计算机技术

项目作者： 杨毅

作者单位： 北京理工大学

项目金额： 85万元

中文摘要： 越野环境中的自主导航是地面无人平台研究领域的一项巨大挑战，本项目以地面无人平台越野自主行驶需求为牵引，针对越野环境的复杂多样和高度非结构特点，拟开展基于图像信息与深度信息融合的越野环境探测、认知及运动规划研究。本项目基于度量空间、语义空间与行为空间的度量-语义-行为的行驶空间架构，研究工作拟主要针对以下四个方面展开：（1）基于RGBID综合特征的图像信息与深度信息的空间配准问题；2）基于水平集等高线特征的越野环境6D-SLAM问题；3）基于RGBID信息的越野环境非参数语义描述问题；4）基于非参数学习的地面无人平台越野环境行驶规划问题。基于以上内容的研究成果，提高地面无人平台在越野环境中的环境认知、描述的精度与速度，提高越野环境下地面无人平台的自主运动规划能力，增强地面无人平台对越野环境的适应性。

中文关键词： 地面无人平台；越野；信息融合；行驶空间；自主导航

英文摘要： Autonomous off-road navigation continues to pose a big challenge to the Unmanned Ground Vehicles community. As to problems of environment perception and autonomous navigation of UGVs, this project intends to research on detecting, perceiving and motion planning under off-road environment based on the combination of image information with depth information. We introduce the concept of metric-semantic-behavioral travel space based on metric space, semantic space and behavior space. On the basis of travel space configuration, our research includes the following parts: (1) space registration of image and depth information based on RGBID comprehensive feature; (2) 6D-SLAM of off-road environment based on Level Set Contour; (3) nonparametric semantic description under off-road environment based on RGBID; (4) motion planning of UGVs under off-road environment based on nonparametric learning. Based on the results of our research, we hope to enhance accuracy and speed of environment perception and description of UGVs under off-road environment, and to improve the performance of autonomous motion planning of UGVs. Besides, we expect the adaptability of UGVs could be robust with our research.

英文关键词： Unmanned Ground Vehicles;Off-road;Information Fusion;Travel Space Configuration;Autonomous Navigation