Research On Mining Scene Localization Technology Based On Multi-sensor Fusion

With the country’s vigorous promotion of manufacturing intelligence,China’s intelligent equipment manufacturing industry is steadily upgrading,and the mining operation scene is faced with a harsh operating environment,heavy operating tasks and complex operating scenarios.In order to save operators’ labour and improve operational efficiency and safety,it is important to introduce intelligent operation technology in the mining scenario.Positioning is a core aspect of construction machinery intelligence.In non-structural scenarios,the fusion of multiple heterogenous sensors has become a research trend as long-term reliable data cannot be obtained using a single sensor.In non-structural environment scenarios,the selection and application of sensor types,arrangement and management,and the design of data fusion schemes remain problematic according to different needs.The resolution of these issues is of great engineering application for accurate positioning in mining scenarios.This thesis relies on the industry-academia-research university-enterprise cooperation project "key technology development for autonomous operation of hydraulic excavator",and takes hydraulic excavator as the research object to explore the positioning technology in the autonomous operation target of excavator.A combination of positioning solutions suitable for mining operation scenarios is studied,and a suitable positioning method system under special working conditions is formed.(1)The classification of mine operation scenes is proposed.According to the different working environment,operation objects and targets of construction machinery,the mine operation scenes are divided into two categories: global scenes and local scenes,and the autonomous positioning requirements under the mine scenes are clarified according to the operation characteristics of different scenes to determine the specific multi-sensor fusion positioning scheme.(2)Common application schemes of multi-sensor data fusion techniques in the field of localisation are investigated.The theories and frameworks for the selection and combination of different sensors,state estimation and data fusion,algorithmic principles of combined localisation techniques,and their application in mining scenarios are discussed.(3)Simulations and principle experiments are carried out on different sensor fusion positioning algorithms.A Kalman filter model is built for the fusion of GPS and inertial navigation system data for simulation to obtain smoother and more reliable positioning results that meet long-term positioning requirements under the assumption of a linear system.For non-linear systems in real mining scenarios,fusion positioning is performed with an error-state Kalman filter algorithm under an open-source dataset.For the vision and laser fusion localisation method,data pre-processing is carried out for the characteristics of the mining scene,2D image and 3D point cloud data features are extracted,and feature fusion matching is carried out by projection transformation to obtain more reliable system state estimation results.Taking intelligent operation of construction machinery as the starting point,this thesis researches multi-sensor fusion positioning technology that meets different positioning needs in mining operation scenarios,in view of the practical problems of insufficient reliability and accuracy of long-term autonomous positioning in nonstructural environments.According to the positioning needs under different intelligent operation scenarios,corresponding sensor combinations and corresponding data fusion schemes are proposed,and data bases are provided for other intelligent modules such as sensing,decision making,planning and control of construction machinery and equipment,resulting in a mining scenario fusion positioning scheme that has important engineering practical significance and application value in promoting intelligent research of construction machinery.