Research On Autonomous Lane Changing Control Strategy Based On Multi-sensor Fusion

Based on the increasingly significant proportion of the automobile and its related industries on the world economic map,the role of the automobile industry in the further prosperity of economy and the iterative renewal of society cannot be ignored.In the face of the complex social environment,the influence of various factors has prompted the automotive industry to enter a comprehensive and in-depth new transformation period.As a result,people's demand for intelligence has spontaneously occurred.At present,the autonomous vehicle has become the strategic commanding height of the global automobile industry's development.The autonomous vehicle has standards for driving safety,which require the sensors of the vehicle perception layer to sense the surrounding environment information quickly,completely and accurately.For this reason,the issues of compatibility and confidence among multiple sensors have received widespread attention,and the heterogeneous multisensor fusion technology arises at the historic moment.Vehicle lane changing behavior is extremely common in daily traffic conditions.The number of accidents in the lane changing process and the corresponding traffic delays caused by human factors bring huge losses.To this end,the vehicle's autonomous lane changing function is developed to improve the vehicle's driving safety,operating efficiency,and environmental adaptability with the support of complete environmental awareness,adaptive decision-making strategies,and precise control algorithms.This paper mainly studies the effective and applicable autonomous lane changing control strategy based on multi-sensor fusion,which includes the perception layer,decision-making and planning layer,control and execution layer in the research of intelligent driving.The research mainly solves two problems: the design of multi-sensor fusion strategy,the research of decision definition scheme,trajectory planning scheme and tracking control scheme of intelligent vehicle's autonomous lane changing behavior.The specific research contents are as follows:Study the definition of decision mode and the scheme of trajectory planning of the intelligent vehicle's autonomous lane changing function.Reasonably completed the simplified process to define the lane change scenario on which the study was based.The concept of satisfaction was used to analyze the generation of lane changing intent,and the feasibility of lane changing was explored based on the minimum safety distance model.When the intention and feasibility of lane change are available at the same time,a lane change decision can be made.Based on the general model of quintic polynomials,the trajectory cluster of autonomous lane changing process was generated.The correlation between the speed of the host vehicle,the time required for lane change,and the efficiency and comfort was analyzed in the MATLAB simulation environment.Given the constraints and cost function,the optimal lane change trajectory was selected by applying the optimization solution theory to weigh the efficiency,safety and comfort of the autonomous lane changing process.Study the trajectory tracking control of intelligent vehicle's lane changing process based on model predictive control.The vehicle monorail dynamics model was chosen as the prediction model of predictive control,which was simplified based on the tire's linearization and small angle assumptions.And the methods of approximate linearization and first-order difference quotient were used to obtain the final discrete linearized prediction model.The increment of the control amount was selected as the system's input,and the state of the system control time domain and predicted time domain was derived based on the current state of the system.The objective function was defined and transformed into a standard quadratic form,related constraints were also defined.The model predictive control process was transformed into an optimization problem of the objective function under constraints in each cycle,and solved by quadratic programming.The most front-end quantity of the optimal control increment sequence was applied to the system,and the solution process of quadratic programming was repeated at each new cycle to complete the track tracking process under the model predictive control.Combined with the vehicle lane changing trajectory of quintic polynomial mode,the simulation test was performed in the Matlab/Simulink environment,which confirmed that the controller has universally effective and stable tracking and control functions.Study on multi-sensor fusion technology.Analyze the characteristic information of sensors commonly used to sense the surrounding environment of the vehicle.Based on the typical sensor layout scheme for intelligent vehicles,combined with the information list required for autonomous lane changing,millimeter wave radar,camera,and rear and side detection radar were selected as the research objects.And their layout on the vehicle was completed.Study the deep meaning and prerequisites of multi-sensor fusion.Combined with the requirements and sensor layout,the comprehensive research scheme of decision-level fusion of rear and side detection radar and feature-level fusion of radar and camera was selected.Feature level fusion used Kalman filter theory and global nearest neighbor data association algorithm to match the observations and targets through a cost matrix.Design the judgment logic located at the fusion center to process the components that cannot be successfully matched to finally obtain the exact target state.Based on the MATLAB automatic driving system toolbox,the simulation experiment of radar and camera sensor fusion strategy was performed.The test results proved that the multi-sensor fusion is more accurate than single sensor detection.A Simulink/PreScan/CarSim joint simulation platform was set up,characteristic conditions were designed,and the simulation experiment of the autonomous lane changing trajectory tracking control was completed.The analysis results showed that the designed decision-making algorithm can reasonably generate the lane changing intention and judge the safety of the lane changing.And no matter what the working conditions were,the trajectory tracking controller designed for autonomous lane changing can follow the ideal lane change trajectory obtained through trajectory planning in a timely and stable manner.The results of simulation tests proved the safety,practicability and reliability of the research results of this paper.Finally,this paper analyzed the shortcomings of the existing research content and gave some solutions.At the same time,combined with the development trend of future automobiles,it pointed out the direction of further study and research.