CN111532274A - Intelligent vehicle lane change auxiliary system and method based on multi-sensor data fusion - Google Patents

Abstract

The present invention proposes an intelligent vehicle lane-changing assistance system and method based on multi-sensor data fusion. 77GHz millimeter-wave radar sensors are installed respectively, and 24GHz millimeter-wave radar sensors are installed on the right front and right rear respectively. The detection ranges of millimeter-wave radar sensors with different frequencies installed in the front and rear of the car will have certain overlapping areas and non-overlapping areas. Firstly, the primary selection of surrounding vehicle targets is carried out for each sensor, and then the validity of surrounding vehicle targets is tested according to the principle of Kalman filtering. Finally, according to the advantages of the sensing range of millimeter-wave radar sensors of different frequencies, the D-S evidence theory is used for the overlapping area. Perform decision-level signal integration. The invention fully integrates the detection advantages and characteristics of each sensor, comprehensively detects the surrounding environment of the vehicle, and timely judges whether the lane change is in a dangerous state.

Description

Intelligent vehicle lane change assistance system and method based on multi-sensor data fusion

technical field

The invention belongs to the field of automobile intelligent and safe driving assistance, and in particular relates to an automobile intelligent lane-changing assistance method for peripheral driving environment monitoring based on multi-sensor data fusion.

Background technique

The intelligent vehicle system is a comprehensive system that integrates functions such as environmental perception, planning decision-making, and multi-level assisted driving. It is a typical, multi-disciplinary, and comprehensive combination of high-tech and high-tech. There is a greater risk of rapidly changing lanes. If the driver underestimates the position and relative speed of the surrounding vehicles when changing lanes, it is very likely to cause the driver to misjudge and change lanes incorrectly, which may cause unnecessary traffic accidents, property damage or casualties. Therefore, the intelligent driver assistance system (ADAS) came into being, which can accurately measure the dangerous vehicles in the blind area of the rearview mirror on both sides of the car, and monitor the vehicles within a certain distance behind the car accordingly.

At present, a large number of researchers at home and abroad have mainly conducted a sufficient analysis of the feasibility of lane changing on the relative motion relationship between the ego vehicle and the vehicle in the blind spot of the rearview mirror and the vehicle behind the target lane at a certain distance (within 60 meters) during the lane changing process. For example, the blind spot monitoring system of Toyota, Land Rover and BMW series, or the automatic driving system of Audi series, its technology has been relatively mature and reliable. However, the above systems basically have no monitoring capability for dangerous vehicles slowing down and slowing down in front of the target lane at a long distance (above 60 meters) and dangerous vehicles accelerating fast behind the target lane at a long distance (above 60 meters). If the driver encounters the above two situations on the highway, and the driver himself has no accurate judgment, in this case forcibly overtaking and changing lanes, especially when both vehicles are in a fast driving state, it is very likely to cause serious traffic accidents . In addition, in order to ensure hardware redundancy in the previous lane-changing assistance system, dual 24GHz millimeter-wave radars are often used as sensor input devices to perform data fusion on dual radars of the same type. The measurement errors and measurement limitations of radars of the same type are often consistent. Data fusion often does not get more accurate results, so the lane change assist system has some drawbacks.

SUMMARY OF THE INVENTION

Aiming at the disadvantages of the existing lane-changing assistance system proposed above, in order to overcome the above shortcomings, the present invention is based on the 77GHz and 24GHz millimeter-wave radar sensors at the front and rear of the vehicle, and adopts corresponding data processing and data fusion means to give full play to the advantages of different frequency sensors. The long-distance (more than 60 meters) dangerous vehicles in the target lane are monitored and early warning; at the same time, the decision-level data fusion processing is carried out on the monitoring overlapping areas between different frequency radars, and more accurate monitoring and early warning results are obtained, which is more accurate and comprehensive. The safety of the lane change assist system is improved, and the possibility of traffic accidents is reduced to a greater extent.

In order to achieve the above-mentioned technical purpose, the technical scheme of the present invention is: a data fusion lane-changing assistance method based on intelligent vehicle multi-sensor, which comprises the following steps: 1) changing the sensors and layout scheme of the existing lane-changing assistance system, the described The intelligent lane-changing auxiliary system includes a data acquisition unit, a signal amplification unit, an independent data processing unit, a data fusion control unit and a warning and early warning unit; 2) The data acquisition unit simultaneously collects the long-distance (0-120 meters) left front and rear of the vehicle. The driving data of the vehicle in the target lane on the left side and the vehicle in this lane, as well as the driving data of the vehicle in the front and rear on the right side (within 60 meters), and the data signal is amplified and transmitted to the independent data processing unit through the CAN bus; 3) All The independent data processing unit respectively preprocesses the collected data, carries out the preliminary selection of the target vehicle and the target validity test, and first generally judges whether it is a dangerous vehicle; 4) The data fusion control unit accepts the signal of the independent data processing unit, and the Monitor the overlapping area for decision-level data fusion processing, and independently judge the non-overlapping area, generate early warning control instructions and transmit them to the warning early warning unit; 5) The warning early warning unit performs corresponding sound and light and steering intervention early warning according to the received signals.

The sensors in the above data acquisition unit are two 77GHz millimeter wave radar sensors and two 24GHz millimeter wave radar sensors.

The installation position and installation layout of the above sensors are as follows: the 77GHz millimeter-wave radar sensors with long detection distance and narrow detection area are installed in the left front and left rear of the car, respectively, and the detectable sensors are installed in the right front and right rear of the car respectively. A 24GHz millimeter-wave radar sensor with a short distance and a wide detection area.

The target lane mentioned above is mainly the left lane of the vehicle. Because the state stipulates that the highway should be overtaking on the left side, the vehicle information in the left lane is mainly detected.

A lane-changing assistance method for intelligent vehicle multi-sensor data fusion is realized by the above system, and the specific steps are as follows:

S1: Collect environmental information within 120 meters before and after the current lane and the target lane through the left front and left rear 77GHz sensors, and collect the front and rear environmental information of the vehicle through the right front and right rear 24GHz sensors;

S2: Independently preprocess the collected data of the four sensors in the front and rear of the car to achieve the primary selection of the vehicle target;

S3: The target validity test based on Kalman filtering is independently performed on the vehicle targets initially selected by the four millimeter-wave radar sensors;

S4: Independently judge the non-overlapping area in front of or behind the car, and determine whether there is a slow-moving vehicle in the long distance in front of the target lane and whether there is an accelerating fast-moving vehicle in the long distance behind;

S5: The D-S evidence theory is used to fuse the data of the two sensors in the front and the rear of the car to monitor the overlapped area, that is, the two independent evidences generated after the target validity test are used to ensure the reliability of the detection to the greatest extent;

S6: Send the result processed by the data fusion control unit to the warning and early warning unit;

S7: The warning and early warning unit performs sound, flashing and steering reverse intervention warning according to the judgment result.

等。The environmental information collected by the sensor in the above step S1 includes: the relative speed v _r , the relative distance x _r and the relative angle between each target vehicle and the vehicle

Wait.

The main methods of performing independent preprocessing on the sensor data in the above step S2 include filtering out empty signal targets, filtering out stationary targets and filtering out false targets.

The main methods for validating the target in the above step S3 are as follows:

其中x_n,j,v_n,j,

分别代表第n个周期内测量到的有效车辆目标的纵向相对距离、相对速度和相对加速度。下一周期的目标车辆状态预测，如下所示：Due to the uncertainty of millimeter-wave radar detection and the existence of errors in the detection method, the validity of the target should be further tested. Here, the Kalman filter prediction method is used to test the validity of the vehicle target. Here, the high-order third-order Kalman filter method is used to predict the effective vehicle target information in the cycle, assuming the state

where x _n,j ,v _n,j ,

respectively represent the longitudinal relative distance, relative velocity and relative acceleration of the effective vehicle target measured in the nth cycle. The target vehicle state prediction for the next cycle is as follows:

分别代表第n+1周期内目标车辆预测值的纵向相对距离、相对速度和相对加速度。对第n+1周期的初选车辆目标信息与上式所得到第n个周期的有效车辆目标信息预测值进行比较验证，比较准则如下：In the above formula, T is the scanning period of the millimeter wave radar, the value is 0.04s, x _(n+1)|n , v _(n+1)|n

respectively represent the longitudinal relative distance, relative speed and relative acceleration of the predicted value of the target vehicle in the n+1th cycle. Compare and verify the primary vehicle target information of the n+1th cycle and the predicted value of the valid vehicle target information of the nth cycle obtained by the above formula. The comparison criteria are as follows:

|x _n+1 -x _(n+1)|n |≤|Δx|

|v _n+1 -v _(n+1)|n |≤|Δv|

分别代表第n+1个周期内两车的纵向相对距离、相对速度和相对加速度的采集值；Δx、Δv、

分别代表比较准则的允许最大误差。在本发明中设定 误差为：In the above formula, x _n+1 , v _n+1 ,

respectively represent the collected values of the longitudinal relative distance, relative velocity and relative acceleration of the two vehicles in the n+1th cycle; Δx, Δv,

respectively represent the maximum allowable error of the comparison criterion. In the present invention, the error is set as:

For the primary vehicle target of the n+1th cycle, if it conforms to the above formula, it is considered that the primary vehicle target is consistent with the valid vehicle target of the nth cycle, and the target vehicle information is updated for this, that is, the valid vehicle target of these two cycles. If it is a dangerous vehicle, it is classified as a dangerous vehicle, and if it is a non-hazardous vehicle, it is classified as a non-hazardous vehicle. If the above formula is not satisfied, the target inconsistency needs to be dealt with. For the targets in the monitoring overlapping area, the D-S evidence theory is used to perform data fusion processing to determine whether it is a dangerous vehicle; To ensure safety, that is, if one of the judgment results of the nth cycle and the n+1th cycle of the non-coincident area is a dangerous vehicle and the other is a non-dangerous vehicle, it is determined as a dangerous vehicle.

Further, the fusion process in step S5 is carried out by adopting the following steps:

a. Target synthesis: The observation results of two independent different frequency sensors in the overlapping area are synthesized into a total output result (ID); D-S evidence theory will derive a basic probability distribution function for the evidence sources from two independent sensors, and then D-S The Dempster combination rule in evidence theory can calculate a new basic probability distribution function that reflects the fusion information generated by the joint action of the two evidences, and according to this probability distribution function, a total output result is synthesized, that is, whether to determine whether there is a dangerous vehicle target. ;

b. Target inference: The present invention will logically generate a target vehicle report with a certain degree of confidence according to a certain degree of reliability, to obtain the observation result of the sensor and make inference, and expand the observation result of the sensor into the target vehicle report; according to the D-S evidence theory to generate The probability distribution function of , will form the target vehicle report with a certain degree of confidence, that is, whether the dangerous vehicle actually exists, and then infer whether it is indeed a dangerous vehicle;

c. Target update: Two different frequency sensors generally have random errors themselves, so a set of reports produced by two time-independent sensors of the same frequency is more reliable than the report produced by either sensor. Therefore, before inference and synthesis of two different frequency sensors, the observation data of the respective frequency sensors are combined (updated); that is, for the primary vehicle target of the n+1th cycle, if the set error is met, the primary vehicle target is considered to be selected. Consistent with the valid vehicle target of the nth cycle, the target vehicle information is updated for this.

As a further step, the intelligent provision of lane change information when changing roads in step S7 is to remind the driver through corresponding sound, light or steering intervention warning, and determine whether the vehicle is in a dangerous state according to the relative distance x _r and relative speed v _r between the vehicles , if it is in a safe state, it will display a green light, and the buzzer will not alarm; if it is in a warning state, it will change to a flashing yellow light, and the buzzer will sound once every two seconds; if it is in a dangerous state, it will turn into a red light state, The buzzer sounds once per second; if the driver fails to see or hear the prompt information of lane change assistance in time, and then forcibly changes lanes, in addition to the sound or flashing alarm, it will also be connected to the power steering module. The steering interface feedback unit performs corresponding torque intervention on the steering, timely reminding the driver not to change lanes at this time.

The present invention has the following outstanding advantages due to the adoption of the above double-layer guarantee technical scheme: 1) the present invention adopts a combination of 24GHz and 77GHz frequency millimeter-wave radar sensors, and successfully decelerates and slows down the vehicle at a long distance in the front of the target lane and the rear of the target lane at a long distance. Vehicles that are accelerating fast are monitored. The insufficiency of the existing lane-changing auxiliary system is solved, and the safety of lane-changing is improved. 2) The present invention uses two different frequency sensors to collect information, so the collected information is more comprehensive and more fault-tolerant, and solves the problem of the same measurement error and measurement limitations of redundant radar sensors of the same type. 3) Adopt a two-layer guarantee scheme in which the data is first processed independently, and then the decision-level data fusion is carried out, and the monitoring overlapping areas and non-overlapping areas are processed separately, which not only ensures the accuracy of data judgment, but also ensures the timeliness of decision-making . 4) The steering feedback interface is introduced into the warning and early warning module, which effectively avoids the driver's distraction and ignores the reminder of the sound or flashing signal, and provides feedback to the driver through the steering wheel tactile sensation in time. The invention makes up for the deficiencies of the existing lane-changing assistance system and improves the safety of the intelligent driving assistance system.

Description of drawings

The present invention has 5 accompanying drawings:

Figure 1 shows the installation location and monitoring area of the millimeter-wave radar on the car;

Fig. 2 is the working flow chart of the data fusion lane changing auxiliary system of the present invention;

3 is a schematic diagram of a data fusion lane changing auxiliary system module of the present invention;

4 is a circuit diagram of a signal amplification module of the data fusion lane change auxiliary system of the present invention;

FIG. 5 is a circuit diagram of the warning and early warning module of the data fusion lane changing auxiliary system of the present invention.

Detailed ways

The technical solutions of the present invention will be further described in detail below through examples and in conjunction with the accompanying drawings.

The present invention mainly proposes an intelligent vehicle lane change assistance method based on 77GHz and 24GHz radar sensor data fusion. The method mainly detects by installing 77GHz millimeter-wave radar sensors in the left front and left rear of the car and 24GHz millimeter-wave radar sensors in the right front and right rear. The 77GHz millimeter-wave radar sensor has a long effective detection distance, up to 120m, but its detection area has a narrow angle of about 30 degrees; while the 24GHz millimeter-wave radar sensor has a short effective detection distance of only 60m, but its detection area has a wider angle. , is about 130 degrees. The installation location and monitorable area of the millimeter-wave radar are shown in Figure 1. The detection range of millimeter-wave radar sensors with different frequencies will have certain overlapping areas (the yellow important area in Figure 1) and non-overlapping areas. First, the surrounding vehicle targets are initially selected for each sensor, and then the surrounding vehicle targets are selected according to the principle of Kalman filtering. Validity check. Finally, according to the advantages of the sensing range of millimeter-wave radar sensors of different frequencies, the D-S evidence theory is used for the decision-level signal integration in the overlapping area. Fully integrate the detection advantages and characteristics of each sensor, comprehensively detect the surrounding environment of the car, timely determine whether the lane change is in a dangerous state, and warn the driver through corresponding sound, light and steering intervention. The signal acquisition and detection processing flow is shown in Figure 2. Show. This sensor arrangement scheme can effectively monitor the information of long-distance dangerous vehicles (the area in the blue line frame in Figure 1), and fuse the information collected by sensors of different frequencies, effectively avoiding the need for data exchange between redundant sensors of the same model. Disadvantages of fusion.

An intelligent vehicle multi-sensor data fusion lane changing auxiliary system mainly includes the following modules: a different frequency millimeter wave radar module for collecting information in real time, an amplifier module for amplifying a dual-channel radar signal, a target verification for the amplified signal, and Data fusion processing radar signal processor module and warning module. The system module diagram is shown in Figure 3. The millimeter-wave radar modules of different frequencies simultaneously collect the long-distance (0-120 meters) left front and left rear target lanes and the driving data of vehicles in this lane, as well as the short-range (within 60 meters) data. ) The driving data of the right front and right rear vehicles are sent to the dual-channel amplifier module for amplification; then the data is transmitted to the processor module through the CAN bus, and the data is subjected to the preliminary selection of the target vehicle, the target validity test and the data. Fusion; the processor module generates an early warning control command and transmits it to the warning early warning module, and the warning early warning module performs corresponding sound and light and steering intervention early warning according to the received signals.

A lane-changing assistance method for intelligent vehicle multi-sensor data fusion is realized by the above system, and the specific steps are as follows:

等；S1: Install a 77GHz millimeter-wave radar sensor with a long detection distance but a narrow detection area in the left front and left rear of the car, respectively, and install a 24GHz millimeter wave with a short detection distance and a wide detection area in the right front and right rear of the car. Radar sensor; the left front and left rear 77GHz sensors collect environmental information within 120 meters before and after the lane and the target lane, and the right front and right rear 24GHz sensors collect environmental information before and after the vehicle. The environmental information collected by the sensors includes : the relative speed v _r , the relative distance x _r and the relative angle of each target vehicle and the own vehicle

Wait;

S2: Amplify the collected data from the four sensors in the front and rear of the car into the amplifier. The schematic diagram of the amplifier is shown in Figure 4. The two sensor signals are IF LC I and IF LC Q respectively. The LMP7716MM/NOPB chip is selected to amplify the signal, and the 3.3V power supply voltage is used to complete the amplification of the sensor signal with the peripheral capacitance and resistance devices. At the same time, the two-stage amplifier structure has higher external gain, better signal-to-noise ratio and higher bandwidth.

S3: Then independently preprocess the data to achieve the primary selection of vehicle targets. The specific implementation steps are as follows:

因此当某个输出满足以上条件时，即可判定为是空信号，从而滤除空信号；a. Filter out empty signal targets: The millimeter-wave radar can track 64 vehicle targets because its own data output is 64 channels. Since no detection targets appear in most cases, there must be a large number of detected empty channels. The data output by the radar signal amplifier after amplification is the default minimum value, that is, the relative speed v _r = 0m/s, the relative distance x _r = 0m, the relative angle

Therefore, when an output satisfies the above conditions, it can be determined as an empty signal, thereby filtering out the empty signal;

b. Filter out stationary targets: In the monitoring environment of millimeter-wave radars with different frequencies, there must be close-to-stationary targets such as guardrails, trees or roadside pedestrians. Filter out stationary objects. The relative angle and relative distance of the stationary target are basically the same as those of the normal dynamic target, but the absolute speed of the stationary target should be 0m/s. Therefore, set the angle between the connection line between the target and the vehicle and the speed direction of the vehicle as α, then the relative speed of the stationary target relative to the vehicle should be close to the vehicle speed, which should satisfy the following equation:

v _r cosα=-v

That is, the absolute value of the stationary target relative to the sum of the speed of the vehicle and the absolute speed of the vehicle should theoretically be equal to 0, but considering that low-speed moving objects such as roadside pedestrians are also considered to be stationary targets for filtering, and taking into account the existence of measurement errors , set the minimum error to 1m/s, and the stationary target that meets the following conditions is filtered out:

|v _r cosα+v|≤1(m/s)

c. Filter out false targets:

False targets refer to those targets that have no objective corresponding or the detected targets appearing for a very short time and have no practical significance, or those targets with poor coherence and large data jumps and fluctuations due to accidental interference of the millimeter-wave radar. This class is a false target. It can be filtered out by the following inequality:

|α _r (n+1)-α _r (n)|≥2°

|x _r (n+1)-x _r (n)|≥4m

|v _r (n+1)-v _r (n)|≥4m/s

Among them, n is the sequence number of the millimeter wave radar sampling point (the description of the motion state of the same target by the radar at different time points), n=(1,2,3,4,5...);

其中x_n,j,v_n,j,

分别代表第n个周期内测量到的有效车辆目标的纵向相对距离、相对速度和相对加速度。下一周期的目标车辆状态预测，如下所示：S4: Carry out the target validity test based on Kalman filtering independently on the vehicle targets initially selected by the four millimeter-wave radar sensors, and use the high-order third-order Kalman filtering method to predict the effective vehicle target information in the cycle, assuming the state

where x _n,j ,v _n,j ,

respectively represent the longitudinal relative distance, relative velocity and relative acceleration of the effective vehicle target measured in the nth cycle. The target vehicle state prediction for the next cycle is as follows:

分别代表第n+1周期内目标车辆预测值的纵向相对距离、相对速度和相对加速度。对第n+1周期的初选车辆目标信息与上式所得到第n个周期的有效车辆目标信息预测值进行比较验证，比较准则如下：In the above formula, T is the scanning period of the millimeter wave radar, the value is 0.04s, x _(n+1)|n , v _(n+1)|n

respectively represent the longitudinal relative distance, relative speed and relative acceleration of the predicted value of the target vehicle in the n+1th cycle. Compare and verify the primary vehicle target information of the n+1th cycle and the predicted value of the valid vehicle target information of the nth cycle obtained by the above formula. The comparison criteria are as follows:

|x _n+1 -x _(n+1)|n |≤|Δx|

|v _n+1 -v _(n+1)|n |≤|Δv|

分别代表第n+1个周期内两车的纵向相对距离、相对速度和相对加速度的采集值；Δx、Δv、

分别代表比较准则的允许最大误差。在本发明中设定 误差为：In the above formula, x _n+1 , v _n+1 ,

respectively represent the collected values of the longitudinal relative distance, relative velocity and relative acceleration of the two vehicles in the n+1th cycle; Δx, Δv,

respectively represent the maximum allowable error of the comparison criterion. In the present invention, the error is set as:

For the primary vehicle target of the n+1th cycle, if it conforms to the above formula, it is considered that the primary vehicle target is consistent with the valid vehicle target of the nth cycle, and the target vehicle information is updated; if the above formula is not satisfied, the Perform target inconsistency processing. For the monitoring of overlapping area targets, the D-S evidence theory is used for data fusion processing and decision. For non-overlapping area targets, the vehicle target is assumed to be valid by default to ensure maximum safety.

S5: Independently judge the non-overlapping area in front of or behind the car to determine whether there is a slow-moving vehicle in the long distance in front of the target lane and whether there is an accelerating fast-moving vehicle in the long distance behind; the independent data processing unit detects the radar sensor data by , judge the speed of the vehicle, and then derive the speed and solve the acceleration to determine whether there is a deceleration or acceleration of the vehicle;

S6: The D-S evidence theory is used to separately monitor the overlapping area of the two sensors in the front and the rear of the car, that is, the two independent evidences generated by the target validity test are data fusion, so as to ensure the reliability of the detection to the greatest extent;

a. Target synthesis: The observation results of two independent different frequency sensors in the overlapping area are synthesized into a total output result (ID);

b. Target inference: the present invention will logically generate a target vehicle report with a certain degree of confidence according to a certain degree of reliability, to obtain the observation result of the sensor and infer, and expand the observation result of the sensor into the target vehicle report;

c. Target update: Two different frequency sensors generally have random errors themselves, so a set of consecutive reports from the same frequency sensor that is sufficiently independent in time is more reliable than any single report. Therefore, the observations of the respective frequency sensors are combined (updated) before inference and synthesis of the two different frequency sensors.

S7: Send the result processed by the data fusion control unit to the warning and early warning unit;

S8: The warning and early warning unit performs sound, flashing and steering reverse intervention warning according to the judgment result. The schematic diagram of warning and warning is shown in Figure 5. The LED_BUSY pin shows whether the chip is busy (during power-on), and the LED_DETECT interface is connected to two external indicators. At the same time, the DETECT_SIGNAL terminal outputs different buzzers according to the level of danger. Response frequency, the DETECT_OUT output signal is similar to DETECT_SIGNAL. Different frequencies are output according to the urgency of the situation. In order to prevent the steering interface ground loop or noise injection, an isolated digital output is used to ensure the stability of the output signal. According to the relative distance x _r and relative speed v _r to determine whether it is in a dangerous state, if it is in a safe state, it will display a green light, and the buzzer will not alarm; if it is in a warning state, it will change to a flashing yellow light, and the buzzer will It will sound once every two seconds; if it is in a dangerous state, it will turn into a red light state, and the buzzer will sound once every second; In addition to sound or flashing alarms, the steering interface feedback unit connected to the steering assist module also performs corresponding torque interventions on the steering through frequency, timely reminding the driver not to change lanes at this time.

The intelligent vehicle multi-sensor data fusion lane change assist system is an intelligent early warning system applied on the highway to prevent collisions when vehicles change lanes. Monitor the distance to dangerous vehicles, and provide accurate lane-changing information and lane-changing danger prompts for vehicles changing lanes. The multi-sensor data fusion of the lane change assist system is to combine the environmental information of the front and rear of the car obtained by two sets of radar sensors with different frequencies, such as the relative speed, relative distance and direction angle of the target vehicle in the far rear and the vehicle. Data collection, data processing, and data fusion are used to analyze and judge the information of target vehicles in the environment, so as to accurately give the degree of danger of changing roads, provide an effective basis for drivers to make lane-changing behaviors, and timely lane-changing danger reminders .

The above is only a preferred embodiment of the present invention, but the protection scope of the present invention is not limited to this. The equivalent replacement or change of the machine inventive concept shall be covered within the protection scope of the present invention.

Claims (9)

1. The utility model provides an intelligent vehicle multisensor data fusion auxiliary system that trades lane which characterized in that includes: the system comprises radars with different frequencies, a signal amplification unit, a data processing unit, a data fusion control unit and a warning and early warning unit, wherein the radars with different frequencies are used for acquiring the front and rear environmental information of the vehicle in real time; the radar signal acquisition areas with different frequencies comprise non-coincident areas and coincident areas.

2. The intelligent vehicle multi-sensor data fusion lane-changing auxiliary system according to claim 1, characterized in that: the radars with different frequencies are a 77GHz millimeter wave radar sensor and a 24GHz millimeter wave radar sensor.

3. The intelligent vehicle multi-sensor data fusion lane-changing auxiliary system according to claim 2, wherein two 77GHz millimeter wave radar sensors are arranged at the left front part and the left rear part of the automobile; the number of the 24GHz millimeter wave radar sensors is two, and the two sensors are arranged at the front right and the rear right of the automobile.

4. The intelligent vehicle multi-sensor data fusion lane change auxiliary method is characterized by comprising the following specific steps:

s1: collecting the environmental information of the front and the back of the lane where the vehicle is located and the target lane by using a 77GHz sensor at the front and the back of the left, and collecting the environmental information of the front and the back of the vehicle by using a 24GHz sensor at the front and the back of the right;

s2: the method comprises the following steps of respectively and independently preprocessing four collected sensor data in front of and behind the automobile to realize the initial selection of a vehicle target;

s3: target validity inspection based on Kalman filtering is independently carried out on vehicle targets preliminarily selected by the four millimeter wave radar sensors;

s4: independently judging the non-coincident region detected in front of or behind the automobile to judge whether a slow-speed vehicle is decelerated in a long distance in front of the target lane and whether an accelerated fast-speed vehicle is accelerated in a long distance behind the target lane;

s5: respectively carrying out data fusion on the superposed areas monitored by the two sensors in front of and behind the automobile, namely two independent evidences which are generated and are subjected to target validity test, by adopting a D-S evidence theory;

s6: transmitting the processed results of the steps S4 and S5 to a warning and early warning unit;

s7: and the warning early warning unit carries out sound, flash and steering reverse intervention early warning according to the judgment result.

5. The intelligent vehicle multi-sensor data fusion lane-changing auxiliary method according to claim 4, wherein the environmental information in the step S1 comprises: relative velocity v of each target vehicle and the host vehicle_rRelative distance, relative distancex_rAnd relative angle

6. The intelligent vehicle multi-sensor data fusion lane-changing auxiliary method according to claim 4, wherein the preprocessing in the step S2 is realized by filtering empty signal targets, filtering static targets and filtering false targets.

7. The intelligent vehicle multi-sensor data fusion lane-changing auxiliary method according to claim 4, wherein the step S3 comprises:

step S3.1: hypothetical states

Wherein x_n,j,v_n,j,

Respectively representing the longitudinal relative distance, the relative speed and the relative acceleration of the effective vehicle target measured in the nth period;

step S3.2: the target vehicle state for the next cycle is predicted as follows:

in the above formula, T is the scanning period of the millimeter wave radar, x_(n+1)|n、、v_(n+1)|n

Respectively representing the longitudinal relative distance, the relative speed and the relative acceleration of the predicted value of the target vehicle in the (n + 1) th period;

step S3.3: comparing and verifying the initially selected vehicle target information of the (n + 1) th period with the effective vehicle target information predicted value of the nth period obtained by the formula, wherein the comparison criterion is as follows:

|x_n+1-x_(n+1)|n|≤|Δx|

|v_n+1-v_(n+1)|n|≤|Δv|

in the above formula, x_n+1、、v_n+1

Respectively representing the longitudinal relative distance, the relative speed and the relative acceleration of the two vehicles in the (n + 1) th period; Δ x,. DELTA.v

Respectively representing the maximum error allowed for the comparison criterion;

step S3.4: for the primary vehicle target in the n +1 th period, if the formula (2) is satisfied, the primary vehicle target is considered to be consistent with the effective vehicle target in the n th period, target vehicle information is updated on the target vehicle, otherwise, target inconsistency processing is performed, and for the target in the monitored overlapping area, data fusion processing is performed by using a D-S evidence theory to determine whether the target is a dangerous vehicle; for the non-coincident region target, the vehicle target is assumed to be valid by default, and the safety is guaranteed to the maximum extent, namely if the results of the judgment of the nth period and the (n + 1) th period of the non-coincident region are that one is a dangerous vehicle and the other is a non-dangerous vehicle, the non-coincident region is determined to be a dangerous vehicle.

8. The intelligent vehicle multi-sensor data fusion lane-changing auxiliary method according to any one of claims 4-7, wherein the fusion process in step S5 is performed by adopting the following steps:

a. target synthesis: the D-S evidence theory derives basic probability distribution functions for evidence sources from two independent sensors, then a Dempster combination rule in the D-S evidence theory calculates a new basic probability distribution function which reflects fusion information and is generated by the joint action of the two evidences, and a total output result is synthesized according to the probability distribution function, namely whether a dangerous vehicle target is determined or not;

b. target inference: according to a probability distribution function generated by a D-S evidence theory, a target vehicle report with certain confidence coefficient is formed, namely whether a dangerous vehicle really exists or not is judged, and whether the dangerous vehicle really exists or not is judged;

c. target updating: the two different frequency sensors themselves are typically randomly error-prone, so that a set of reports generated by the two time-independent same frequency sensors is more reliable than a report generated by either sensor. Thus, prior to inferencing and synthesis of two different frequency sensors, the observed data of the respective frequency sensors are combined.

9. The intelligent vehicle multi-sensor data fusion lane change assisting method according to claim 4, wherein when the road is changed in the step S7, the intelligently provided lane change information reminds a driver through corresponding sound, light or steering intervention early warning. According to the relative distance x between vehicles_rAnd relative velocity v_rJudging whether the vehicle is in a dangerous state, if so, displaying a green lamp, and not giving an alarm by the buzzer; if the alarm is in the warning state, the alarm is changed into a flashing yellow light, and the buzzer rings once every two seconds; if the alarm is in a dangerous state, the alarm is converted into a red light state, and the buzzer rings once per second. If the driver does not see or hear the prompting information of lane change assistance in time, and then the lane change is forcibly carried out, besides sound or flash alarm, corresponding torque intervention is carried out on the steering through a steering interface feedback unit connected into the power steering module, and the driver is timely reminded not to make a lane change behavior at the moment.

Images