Research In Identifying Of Vehicles’ Relative Position Based On Millimeter-Wave Radar

Millimeter-wave radar is usually used in vehicle active systems to identify theexistence of other vehicles around. However the data output by the millimeter-wave radar isalways mixed with wrong signals which disturb the normal work of vehicle active systems.In addition, the data of vehicles identified by the millimeter-wave radar can’t be used directto vehicle active safety system because the relative position between other vehiclesobjectives and own vehicle can not be read from which.In order to solve the problems mentioned above, we use ESR millimeter-wave radar tocollect the data in the actual road tests and build the algorithm to delete the wrongobjectives. On the other side, the road curvature is estimated by the yaw rate and velocity ofown vehicle. With the road curvature data, the distance between own vehicle and lane andthe other vehicles’ data output by millimeter-wave radar, we establish different models instraight and curved sections，which can be used to identify the relative position state of theother vehicles to own vehicle. And the models can provide precise vehicles’ information.The main contents and conclusion are as follows:1. An effective vehicle target filter algorithm is designed. As each millimeter-waveradar can collect64units’ data information, there exists a lot of non-vehicle data. Theeffective vehicle target filter algorithm can identify the moving vehicles, and provide thedata of which to the relative position state identify model.2. A road curve estimate model based on yaw rate and velocity of own vehicle isestablished. And then the kalman-filter model based on2DOF vehicle model is built toimprove the accuracy of curve estimate.3. Different relative position state identify models in straight and curved sections arebuilt. Which include the identifying when the vehicle target is in own vehicle’s same lane,neighbor lane or non-neighbor lane.4. The verification of relative position state identify model is accomplished by the useof test data and pictures. Different identify models are verified by the data and pictures fromfront and rear areas. The verification results show that the different relative position state identify models can work effectively in straight and curved sections. And which can provideeffective data of vehicles to the control system.