Research On Intelligent Control Of Unmanned Vehicle Based On Sensor Occluded Scenes

Rapid development of the automobile industry has improved the efficiency of people’s travel.And the emergence of unmanned vehicles has already improved life quality of people with the aim of liberating drivers.However,frequent occurrences of automobiles collision accidents have also brought huge security risks and economic losses to the people due to technical limitations and complexity driving environment.Three major links of driverless cars are crucial.The environment perception link identifies road environment,the planning decision-making link carries out dangerous judgments on the environment in which the vehicle is located and controls the execution link to brake or turn the vehicle.Due to the limitations and effects of the environmental sensing sensors,they will not recognize obstacles that are completely obscured by trees,mountains,etc.,which poses a great safety hazard.To this end,the above-mentioned environmental areas with great security risks are called potential traffic accident-prone areas,that is,sensor occluded scenes.Data mining algorithms are used to deeply mine characteristics of sensor occluded scenes,providing more real-time and detailed data for the control strategy of driverless vehicles.In this paper,the relevant characteristics of sensor occluded scenes are deeply analyzed,and prediction model of potential obstacles in occluded scenes is established.Based on the static obstacles in potential traffic accident areas,an innovative safety distance model is established.At the same time,the active collision avoidance control algorithm applied in sensor occluded scenes is studied based on possible dynamic obstacles.Firstly,environment-aware sensors used in the field of intelligent driving are introduced;causes of frequent traffic accidents are analyzed;sensor occluded scenes are made in-depth research;the inevitable effect of occluded scenes in the driving process of driverless vehicles is analyzed.Unmanned vehicles cannot effectively prevent collisions if some obstacles exist in sensor occluded scenes because they are not directly observable.Therefore,it is necessary to make reasonable and effective predictions of possible obstacles in sensor occluded scenes.The domestic and international research results of theories and methods involved in this paper are analyzed,and the one-sidedness of the existing research directions is pointed out.The overall scheme of potential traff-ic accident active collision avoidance system is further designed.Then,it is proposed to use the boundary line of the sensor sensing as the location where the sensor may be aware of obstacles in sensor occluded scenes.The convolutional neural network is used to identify the type and location of sensor occluded scenes in real time.According to the standard parameters of vehicle and the driving conditions,the motion trend of sensor occluded scenes are analyzed and prediction model of potential obstacles in occluded scenes is built.This model can better reflect the motion state the sensor occluded scenes motion trend,which provide a good foundation to study potential traffic accident areas active collision avoidance algorithm.Next,parameters such as vehicle speed,perceived blind zone motion state,relative distance between vehicle and perceived blind zone are obtained and the innovative safety distance model is established based on static obstacles that may exist in sensor occluded scenes.Meanwhile,active collision avoidance algorithm is studied based on the possible dynamic obstacles and the simulation experiment is carried out.In MATLAB and CARSIM,simulations are carried out under different working conditions,and compared with the simulation results of the safety distance model without considering sensor occluded scenes.Simulation results show that the innovative safety distance model complements the existing safety distance model and the active collision avoidance algorithm studied greatly improves the active safety performance of the unmanned car.What’s more,it is especially suitable for road environments such as comers and intersections where tall buildings exist.Finally,the selected active collision avoidance controller is subjected to offline testing and real vehicle testing.Experimental results show that the selected sensor occluded scenes active collision avoidance controller can well adapt to a variety of special and emergency conditions,can accurately complete the precise control of the active collision avoidance,and avoid the collision accident to the greatest extent.It also provides a reference for the research of sensor occluded scenes related problems and collision avoidance system.