| In recent years,the national energy security strategy has been upgraded,and the convergence of the new generation information technology and the coal industry has been accelerated.It is extremely urgent to transform and upgrade the high-end,intelligent and green coal industry.Only by adhering to the new development concept to realize the transformation of coal mining and utilization methods can we effectively improve the intelligence and safety level of coal mines and promote the high-quality development of the coal industry The foundation and production conditions of intelligent construction in coal mines in China are complex,diverse,unbalanced and insufficient.The research on automation and intelligence of fully mechanized coal mining face started earlier,and the construction of intelligent coal mining face is in full swing.However,the research on intelligence of underground roadway heading face in coal mines develops slowly due to the variability of geology,environment,technology and supporting equipment,which cannot match the implementation progress of fully mechanized coal mining intelligence.Therefore,this dissertation takes the roadheader,the key equipment of coal mine roadway,as the research object,based on inertial positioning technology and visual positioning technology,aiming at realizing the passive autonomous positioning of roadheader,and carries out the research on key technologies of visual and inertial fusion autonomous positioning of coal mine roadheader.Based on the cutting load of roadheader and its kinematic slip model,this dissertation explores the distribution law and compensation strategy of inertial navigation positioning error of roadheader under complex vibration conditions,optimizes the extraction and matching method of visual positioning features in coal mine roadway environment,and constructs a vision/ineitia fusion autonomous positioning system of roadheader under multi-state constraints,thus realizing the accurate positioning of roadheader in simulated roadway.Specific research contents include:On the basis of kinematics and dynamics sliding model of crawler walking mechanism of roadheader,combined with roadway environment and excavation process,the autonomous positioning strategy of roadheader is put forward.The characteristic load of roadheader is obtained by rock breaking test of roadheader.The motion characteristics of roadheader walking mechanism is analyzed,and its kinematic slip model and dynamic slip model sre established.The posture response of roadheader is obtained.Based on the environmental characteristics and the analysis of tunneling technology,the autonomous positioning strategy of visual/inertial fusion of roadheader and its key technologies is proposed.Based on the principle of inertial navigation system posture update and error transfer,the error distribution law and compensation correction method of inertial posture calculation of roadheader under dynamic constraints are explored.The error parameter model of inertial sensor is analyzed,and a fast calibration device for inertial sensor is designed and developed.Based on the inertial navigation posture update equation,a mileage encoder aided inertial navigation solution model of roadheader based on Kalman filter is established,and a zero-speed soft feedback correction update method is proposed,which conforms to the reciprocating start-stop process of roadheader.The dynamic theory and simulation model of roadheader is established,the error compensation algorithm of inertial navigation positioning of roadheader in complex vibration environnment is explored,and the state estimation models of unscented Kalman filter and cubature Kalman filter are established according to the nonlinear working state of roadheader.The inertial positioning test-bed of roadheader is built to realize the inertial positioning of roadheader and acquire its positioning error accuracy.Based on the principle of binocular vision positioning,this dissertation explores the effective extraction and matching of image features in coal mine roadway,and establishes the visual positioning system of roadheader.On the basis of detailed analysis of the characteristics of roadway visual images,the dust removal and defogging algorithms suitable for coal mine images are comprehensively selected with four evaluation indexes.The pose measurement model of binocular vision is established,and the uncertainty of binocular depth estimation is analyzed.Taking feature quality and matching efficiency as the main indexes,this dissertation studies the roadway visual adaptability of various feature description methods,and proposes a roadway visual image matching method based on mixed features.A visual pose estimation system for simulating roadheader iss built,and the visual positioning accuracy of roadways under conventional illumination and low illumination is tested,which verifies the feasibility of the visual pose estimation system based on mixed features in roadway environment.Based on the multi-state constrained Kalman filter,the visual and inertial fusion positioning system model of roadheader is established,and relevant experimental research is carried out.In view of the shortcomings of inertial positioning technology and visual positioning technology in roadheader positioning,a visual and inertial fusion model of roadheader is established in the form of errors of inertial and visual states,and the prediction propagation and observation update equations of inertial state and visual state are deduced.Accurate coordinate estimation of mature feature points is carried out in the form of inverse depth,and a binocular visual inertial positioning system of roadheader assisted by mileage encoder and its update mechanism is designed.The co-calibration method of inertial sensor and visual sensor based on gravity vector is put forward.The visual inertial fusion positioning test system of simulated roadheader is built to realize the accurate autonomous positioning of simulated roadheader under conventional driving mileage and long-distance complex path,which provide a new method and technical support for intelligent application of coal mine roadheader.The dissertation has 116 figures,16 tables and 163 references. |
PDF Full Text Request