Font Size: a A A

Intelligent Fusion Positioning Technology For Vehicles In Urban Environment

Posted on:2019-11-08Degree:DoctorType:Dissertation
Country:ChinaCandidate:Q M XuFull Text:PDF
GTID:1362330590960105Subject:Instrument Science and Technology
Abstract/Summary:PDF Full Text Request
In order to meet the requirement of the new-generation Intelligent Transportation System(ITS)for vehicle positioning in complex urban environment,an important trend is developing a multi-sensor fusion approach for vehicle localization.Based on the idea of multi-sensor fusion,this dissertation proposes an intelligent fusion positioning strategy to improve the accuracy and reliability of vehicle localization in complex urban environment from both aspects of information acquisition and system fusion.First,the key parameters required by vehicle positioning are accurately estimated and the navigation parameters resolving method of reduced MEMS-INS is improved,thus accurate information from multi-sensor can be obtained to provide foundations for fusion positioning.Further,the efficient robust fusion algorithm and hybrid intelligent compensation model for the accumulated position errors are proposed to execute the fusion of multi-sensor information,and therefore improving the whole performance.The results of comprehensive validation experiments indicate that the proposed intelligent fusion positioning strategy can achieve good performance in complex urban environment.The main contribution and innovation of this dissertation can be summarized as:1)Intelligent estimation and multi-parameter joint estimation methodologies are proposed to improve the accuracy of the key parameters estimation required by vehicle localization using low-cost onboard sensors.First,different from the conventional estimation method,an intelligent estimation methodology is proposed for heading angle.An intelligent heading information perception module is developed based on deep learning algorithm to excavate the heading angle related information.An enhanced digital map is utilized to obtain the heading angle of the road.The heading angle of the vehicle can be obtained by combining the heading angle related information and the heading angle of the road.Besides,through combining the kinematics and dynamics model,the joint estimation of longitudinal velocity,lateral velocity,and yaw rate can be achieved by developing a observer based on UKF.The two proposed estimation approaches can address the problem of accurate heading angle estimation in urban environment and improve the accuracy of longitudinal velocity,lateral velocity,and yaw rate estimation.To the author's knowledge,the proposed heading angle estimation approach based on deep learning and enhanced digital map is seldom evaluated or discussed.The results of simulation and field tests indicate that the two estimation approaches mentioned above can improve the accuracy of corresponding parameters to a certain extent.2)An improved navigation parameter resolving method is proposed to deal with the problem that conventional reduced MEMS-INS cannot resolve the navigation parameters completely and accurately.First,the pitch and roll angle are estimated based on kinematics model and the effect of MEMS inertial sensor noise is eliminated.The estimated pitch and roll angle can be used to replace the measurements of the two removed gyros.Then the yaw rate estimated by the multi-parameter joint estimation method is utilized to resolve heading angle.Thus,the accuracy of the attitude angles can be improved.Further,the estimated pitch and roll angle are utilized to correct the vertical acceleration.Then,the modified vertical acceleration and the horizontal accelerometers are used to obtain the velocities and positions of the vehicle.Thus,the navigation parameters can be resolved completely and accurately.According to the experimental results,for the MEMS inertial sensors used during the experiments,the attitude angles and horizontal velocities resolved by the modified method have higher accuracy than those resolved by the conventional method,the accumulated errors in the position resolved by the improved method are also superior to those of the conventional method.3)An interacting multiple model based sequential two-stage Kalman filter(IMM-STSKF)is proposed to enhance the robustness of the fusion algorithm for MEMS inertial sensor noise.Different models of the error variables are established for different noise levels and fused by IMM algorithm.Thus the effect of MEMS inertial sensor noise can be reduced by adaptively adjusting the filter parameters according to the actual noise status.On this basis,a two-stage filtering structure can be formed by combining the IMM based bias filter with the sequential measurement update based bias-free filter.Thus,the proposed fusion algorithm can improve the robustness and maintain the computational cost at a low level simultaneously.From the experimental results,it can be concluded that the proposed IMM-STSKF algorithm has relatively strong robustness for MEMS inertial sensor noise and can improve the positioning accuracy to a certain extent.4)A hybrid intelligent compensation model for accumulated position errors is proposed to address the problem that conventional model cannot accurately describe the accumulated errors.First,the accumulated errors are decoupled as primary accumulation error and uncertain noise related error.Then,the primary accumulation error is online modeled by time series analysis method,while the uncertain noise related error is offline trained by neural network method.The hybrid intelligent compensation model is a combination of the two model established above.The experimental results indicate that the error predicted by the hybrid intelligent compensation model is more accurate than that predicted by conventional model.Thus,the vehicle localization performance can be further improved when satellite positioning is unavailable.
Keywords/Search Tags:vehicle positioning, multi-sensor fusion, parameter estimation, positioning error compensation, artificial intelligence algorithm, reduced MEMS-INS
PDF Full Text Request
Related items