巡线机器人动态时变环境下的特征识别与路径规划研究

Inspection robot for high voltage power transmission lines can improve the safety of power dilivery and the intelligence of the power grid. Currently, most of the inspection robots are controled remotely by operators instead of automatic inspection, because it is difficult to do the feature extraction and motion planning in time-varying environment due to the flexibility of the power transmission line. In this project, we will study the key issues of environment medeling and motion planning of automatic inspection, such as irregular motion image deblurring, time-varying feature extraction on complex background, modeling of dynamic and flexible environment and path plannning without collision. A motion image restoration algorithm based on modeling of image blurring process will be developed to sovle the image blurring problem caused by the irregular camera shaking due to the robot motion along the flexible conductor. A feature extraction algorithm based on deformed shape with variable parameters will be developed to solve the feature recognition problem with complex and flexible environment. An environment modeling method based on static modeling and dynamic estimate will be developed to solve the modeling problem of the dynamic and flexible environment. A rapidly-random-searching -tree based trajectory searching method and a multi-bounding-volume based collision checking method will be developed to sovle the non-collision path planning problem of the inspection robot under time-varying environment with multi-model primitives. Along with our study, including field experiments, a complete solution to automatic power line inspection will be developed, which will lay the foundation for the practical use of the inspection robot.

巡线机器人用于输电线路巡检可以提高电力输送的安全性和电网智能化水平，目前巡线机器人不具备自动巡检能力而只能遥控，难点在于由输电导线柔性带来的时变环境下特征识别和运动规划问题，本课题拟从上述两方面着手，深入探讨无规则运动图像复原、复杂背景下时变参数特征提取、动态柔性环境建模及时变环境下无碰撞路径规划等关键科学问题：提出基于图像模糊过程建模的运动图像复原方法，解决因机器人在柔性导线上运动导致的镜头无规则晃动而出现的图像模糊问题；提出基于可变参数的弹性形状特征提取方法，解决因复杂柔性环境下特征信息时变而难以识别的问题；提出基于静态建模和动态预测相结合的建模方法，解决动态柔性环境建模问题；提出基于快速随机搜索树的路径搜索和基于多包围体碰撞检测相结合的路径规划方法，解决机器人在多模型元素时变环境下的无碰撞路径规划问题。从而结合现场实验形成一整套输电线路自动巡检解决方案，为巡线机器实用化奠定基础。

架空输电线路的安全稳定运行关系对国民经济和社会生活至关重要，因此需要定期巡检以防故障造成重大事故。用巡线机器人代替人工进行线路巡检可提高巡检效率和安全性，而面对复杂动态的输电线路环境，进行环境特征识别和路径规划是巡线机器人自主巡线的前提和关键。本项目主要研究高压输电线路环境特指识别及路径规划问题，通过低分辨率图像下的导线自动检测、复杂背景下绝缘子识别与定位、双目立体视觉下的输电线路近距离三维位姿检测以及基于免模型学习方法的机器人轨迹规划研究，实现了巡线机器人对线路动态环境的实时感知与路径规划。另外还针对架空输电线路典型缺陷的自动检测问题进行了深入研究，实现了导线断股、导线异物悬挂两类典型缺陷的自动检测。通过上述研究，解决了输电线路自动检测的一系列基础和前提问题，为实现了线路自动巡检奠定了基础。