A dynamic path prediction and control method in an intelligent vehicle anti-collision device

Abstract

The present invention discloses a dynamic path prediction and control method in an intelligent vehicle anti-collision device, which relates to the field of intelligent driving technology, and comprises the following steps: step 1: road section marking, step 2: a cloud server determines the road section information; step 3: the cloud server collects the information of oncoming vehicles in the same road section, and makes a judgment based on the traveling information of the oncoming vehicles to give a second-level reminder to the driving vehicle. In the invention: the cloud server performs a risk analysis on the driving section of the driving vehicle to give a first-level reminder to the driving vehicle, and can perform primary dynamic path control on the vehicle; the cloud server and the driving vehicle collect the traveling information of the oncoming vehicles in front of the driving vehicle, and at the same time, the cloud server performs dynamic path prediction on the traveling information of the oncoming vehicles, and then feeds it back to the driving vehicle, so that the driving vehicle can be dynamically predicted and controlled while reducing the workload of the driving vehicle processor.

Description

Technical Field

[0001] The present invention relates to the field of intelligent driving technology, and in particular to a dynamic path prediction and control method in an intelligent vehicle anti-collision device. Background Art

[0002] Intelligent driving essentially involves the cognitive engineering of attention capture and distraction, encompassing three key components: network navigation, autonomous driving, and human intervention. The prerequisite for intelligent driving is that the vehicle we select meets driving dynamics requirements, and that its sensors can acquire relevant visual and auditory signals and information, controlling the corresponding follow-up systems through cognitive computing.

[0003] The main research direction of existing intelligent driving is collision avoidance, and the dynamic path prediction and control of vehicles are the main research topics.

[0004] According to the authorization announcement number: CN106969777B-Path Prediction Device and Path Prediction Method, which records "generating and storing driving path information based on information collected from multiple vehicles including the first vehicle, the driving path information is information associated with the path traveled by each vehicle; a position information acquisition unit, obtaining a first position as the current position of the first vehicle; and a path prediction unit, using at least any one of the first path information representing the path of the first vehicle and the second path information representing the path of the second vehicle in the driving path information stored in the storage unit to predict the driving path of the first vehicle, the second vehicle being a vehicle other than the first vehicle, and the path prediction unit determining the utilization ratio of the first path information used in predicting the driving path based on the past driving record of the first vehicle at the first position." Those skilled in the art can know that this patent performs path prediction by recording information associated with the paths traveled by each vehicle, and lacks a real-time prediction function, resulting in an inability to perform real-time prediction based on the situation of the vehicle ahead.

[0005] According to the authorization announcement number: CN107000746B - System and method for vehicle path prediction, which records "sensing the speed, direction and yaw rate of the vehicle; sensing the steering angle of the vehicle; sensing the driving lane near the vehicle or the driving lane along which the vehicle is traveling; calculating a first path prediction for a first time period after the current time, the first path prediction including a trajectory predicted based on the sensed speed, direction and yaw rate; calculating a second path prediction for a second time period, at least some of the second time period being later than the first time period, where it is assumed that steering actions caused by changes in the steering angle will take effect on the vehicle; calculating a third path prediction for a third time period, at least some of the third time period being later than the second time period, where it is assumed that the driver of the vehicle will control the vehicle's trajectory to attempt to at least generally follow the driving lane; and formulating a combined predicted path for the first time period, the second time period and the third time period." Persons skilled in the art will understand that this patent requires a large budget for sensing the driving speed, direction and yaw rate of the oncoming vehicle, which places a heavy burden on the vehicle system operation, easily leading to system overload and reducing the reliability of intelligent driving.

[0006] In summary, a dynamic path prediction and control method in an intelligent vehicle collision avoidance device is designed. Summary of the Invention

[0007] In order to overcome the above-mentioned deficiencies, the present invention provides a dynamic path prediction and control method in an intelligent vehicle anti-collision device.

[0008] The present invention achieves the above-mentioned purpose through the following technical solutions:

[0009] A method for dynamic path prediction and control in an intelligent vehicle collision avoidance device comprises the following steps:

[0010] Step 1: Road section marking: the driving vehicle connects to the cloud server and marks the road section through the cloud server;

[0011] Step 2: The cloud server determines the road section information and sends the risk level of the road section to the driving vehicle as a first-level reminder;

[0012] Step 3: The cloud server collects information about oncoming vehicles on the same road section and makes a judgment based on the oncoming vehicle's movement information to issue a second-level reminder to the driver;

[0013] The step three comprises the following steps:

[0014] S31. Collecting vehicle information of incoming vehicles through a cloud server, and searching for information of incoming vehicles of the same brand on the same road section and in a common interconnected network;

[0015] S32, collecting oncoming vehicle information by the driving vehicle, and transmitting the oncoming vehicle information in front of the driving vehicle to the cloud server through sensors;

[0016] S33. Determine the route of the oncoming vehicle. The cloud server, based on the collected information about the oncoming vehicle, connects with the vehicle of the same brand and collects the route of the oncoming vehicle of the same brand. If the oncoming vehicle is not of the same brand, the cloud server predicts the path of the oncoming vehicle based on the route information collected by the driving vehicle.

[0017] S34. Perform a second-level reminder. If the driving vehicle and the oncoming vehicle are on a dangerous route, the cloud server sends a second-level reminder to the driving vehicle. If the driving vehicle and the oncoming vehicle are not on a dangerous route, the cloud server continues to collect oncoming vehicle information.

[0018] The step 1 comprises the following steps:

[0019] S11, vehicle positioning: the driving vehicle communicates with the cloud server, which accurately locates the driving vehicle and communicates with the Beidou satellite;

[0020] S12, road section positioning: the cloud server analyzes the road section where the driving vehicle is located based on the vehicle's location and the navigation software;

[0021] S13, road section risk analysis: The cloud server performs a risk analysis on the road section based on the data collected by the driving vehicle sensors and the information collected by the navigation software on the road section;

[0022] S14: Mark the road section. If it is a risky road section, proceed to step 2; if it is not a risky road section, proceed to step S11.

[0023] In step S13, the formula for analyzing the road section risk is:

[0024] ;

[0025] is the real-time risk base value;

[0026] ,in, , calculate the information entropy of each parameter in real time to reflect the impact of data uncertainty on the weight;

[0027] is a nonlinear normalization function,

[0028] The threshold is determined by the quantile of historical data, k is the risk acceleration coefficient,

[0029] To predict the risk increment, is the adjustment coefficient,

[0030] Implementation based on spatiotemporal attention mechanism LSTM:

[0031] , the input includes the risk sequence of the previous 60 minutes, the road network topology matrix , time feature vector .

[0032] As a preference, the cross-influence matrix is ​​added to the formula for the road section risk analysis. The cross-influence matrix is:

[0033] Where i and j are risk factor numbers (e.g., i=1 represents rainfall intensity, j=2 represents emergency braking frequency);

[0034] k is the time window index (every 5 minutes is a slice);

[0035] l is the spatial location code (graph node number based on road network topology),

[0036] The formula for road section risk analysis is as follows:

[0037] .

[0038] Preferably, the , The value is adjustable.

[0039] Preferably, the cloud server predicts the route of the oncoming vehicle through a spatiotemporal trajectory prediction model.

[0040] Preferably, the first-level reminder includes four reminder levels: green, yellow, orange and red, and the second-level reminder is red.

[0041] Preferably, the driving vehicle is equipped with a 128-line laser radar, a 4K surround-view camera and a millimeter-wave radar. The 128-line laser radar is used to scan the steering angle of the vehicle wheel with an accuracy of ±0.5°. The 4K surround-view camera recognizes the flashing pattern of the turn signal, such as three consecutive flashes representing an emergency lane change. The relative speed change trend is calculated through the Doppler effect to predict the acceleration / deceleration trajectory).

[0042] Preferably, the The value of is between 0 and 1. At [0,0.4), the first level reminder is green, and regular monitoring continues. At [0.4,0.6), the first level warning is yellow, and the navigation system prompts "Drive with caution". At [0.6,0.8), the first-level reminder is orange, and a deceleration suggestion is pushed to surrounding vehicles, and a warning message is displayed on the roadside screen. .8, the first-level reminder is red, which automatically triggers lane-level control: the speed limit is reduced by 30%, the anti-collision system is turned on, and a patrol car is dispatched to deal with the situation on the spot.

[0043] The beneficial effects of the present invention are as follows: in the dynamic path prediction and control method in the intelligent vehicle anti-collision device:

[0044] 1. Through the cloud server, the risk analysis of the driving road section is carried out to provide the driver with the first-level warning, and the vehicle can be controlled at the primary level of dynamic path;

[0045] 2. The cloud server and the driving vehicle collect driving information of oncoming vehicles in front of the driving vehicle. At the same time, the cloud server makes a dynamic path prediction of the oncoming vehicle's driving information and then feeds it back to the driving vehicle. This allows for dynamic path prediction and control of the driving vehicle while reducing the workload of the driving vehicle's processor.

[0046] 3. Adding the cross-impact matrix to the road section risk analysis can improve the reliability of the road section risk analysis. BRIEF DESCRIPTION OF THE DRAWINGS

[0047] The present invention will now be described by way of example with reference to the accompanying drawings, in which:

[0048] Figure 1 It is a step diagram of the road section marking of the present invention;

[0049] Figure 2 It is a step diagram of step three of the present invention;

[0050] Figure 3 It is a logic diagram schematic diagram of step three of the present invention. DETAILED DESCRIPTION

[0051] The present invention will now be described in further detail with reference to the accompanying drawings, which are simplified schematic diagrams that illustrate the basic structure of the present invention in a schematic manner.

[0052] like Figure 1-Figure 3 As shown, a dynamic path prediction and control method in an intelligent vehicle collision avoidance device includes the following steps:

[0053] Step 1: Road section marking: the driving vehicle connects to the cloud server and marks the road section through the cloud server;

[0054] Step 2: The cloud server determines the road section information and sends the risk level of the road section to the driving vehicle as a first-level reminder;

[0055] Step 3: The cloud server collects information about oncoming vehicles on the same road section and makes a judgment based on the oncoming vehicle's movement information to issue a second-level reminder to the driver;

[0056] The step three comprises the following steps:

[0057] S31. Collecting vehicle information of incoming vehicles through a cloud server, and searching for information of incoming vehicles of the same brand on the same road section and in a common interconnected network;

[0058] S32, collecting oncoming vehicle information by the driving vehicle, and transmitting the oncoming vehicle information in front of the driving vehicle to the cloud server through sensors;

[0059] S33. Determine the route of the oncoming vehicle. The cloud server, based on the collected information about the oncoming vehicle, connects with the vehicle of the same brand and collects the route of the oncoming vehicle of the same brand. If the oncoming vehicle is not of the same brand, the cloud server predicts the path of the oncoming vehicle based on the route information collected by the driving vehicle.

[0060] S34. Perform a second-level reminder. If the driving vehicle and the oncoming vehicle are on a dangerous route, the cloud server sends a second-level reminder to the driving vehicle. If the driving vehicle and the oncoming vehicle are not on a dangerous route, the cloud server continues to collect oncoming vehicle information.

[0061] The step 1 comprises the following steps:

[0062] S11, vehicle positioning: the driving vehicle communicates with the cloud server, which accurately locates the driving vehicle and communicates with the Beidou satellite;

[0063] S12, road section positioning: the cloud server analyzes the road section where the driving vehicle is located based on the vehicle's location and the navigation software;

[0064] S13, road section risk analysis: The cloud server performs a risk analysis on the road section based on the data collected by the driving vehicle sensors and the information collected by the navigation software on the road section;

[0065] S14: Mark the road section. If it is a risky road section, proceed to step 2; if it is not a risky road section, proceed to step S11.

[0066] In step S13, the formula for analyzing the road section risk is:

[0067] ;

[0068] is the real-time risk base value;

[0069] ,in, , calculate the information entropy of each parameter in real time to reflect the impact of data uncertainty on the weight;

[0070] is a nonlinear normalization function,

[0071] The threshold is determined by the quantile of historical data, k is the risk acceleration coefficient,

[0072] To predict the risk increment, is the adjustment coefficient,

[0073] Implementation based on spatiotemporal attention mechanism LSTM:

[0074] , the input includes the risk sequence of the previous 60 minutes, the road network topology matrix , time feature vector .

[0075] Specifically, the formula for the road section risk analysis is added with a cross-influence matrix, which is:

[0076] Where i and j are risk factor numbers (e.g., i=1 represents rainfall intensity, j=2 represents emergency braking frequency);

[0077] k is the time window index (every 5 minutes is a slice);

[0078] l is the spatial location code (graph node number based on road network topology),

[0079] The formula for road section risk analysis is as follows:

[0080] .

[0081] Specifically, the , The value is adjustable.

[0082] Specifically, the cloud server predicts the route of the oncoming vehicle through a spatiotemporal trajectory prediction model.

[0083] Specifically, the first-level reminder has four reminder levels: green, yellow, orange and red, and the second-level reminder is red.

[0084] Specifically, the driving vehicle is equipped with a 128-line laser radar, a 4K surround-view camera, and a millimeter-wave radar. The 128-line laser radar is used to scan the steering angle of the vehicle's wheels with an accuracy of ±0.5°. The 4K surround-view camera identifies the flashing pattern of the turn signal, such as three consecutive flashes representing an emergency lane change. The relative speed change trend is calculated through the Doppler effect to predict the acceleration / deceleration trajectory).

[0085] Specifically, the The value of is between 0 and 1. At [0,0.4), the first level reminder is green, and regular monitoring continues. At [0.4,0.6), the first level warning is yellow, and the navigation system prompts "Drive with caution". At [0.6,0.8), the first-level reminder is orange, and a deceleration suggestion is pushed to surrounding vehicles, and a warning message is displayed on the roadside screen. .8, the first-level reminder is red, which automatically triggers lane-level control: the speed limit is reduced by 30%, the anti-collision system is turned on, and a patrol car is dispatched to deal with the situation on the spot.

[0086] As Example 1: An example of road section risk analysis is as follows:

[0087] Scene background:

[0088] Geographical features: Long downhill curve (800m radius), historical accident rate 2.3 times higher than the average

[0089] Real-time environment: moderate rain (precipitation intensity 2.8mm / h), visibility 350m, road surface temperature 12℃

[0090] Traffic status: Traffic density: 42 vehicles / minute, large trucks account for 38%

[0091] Special event: A hazardous chemical vehicle was parked in the emergency lane (lasting 9 minutes)

[0092] Calculation process breakdown:

[0093] Data preprocessing

[0094] Road information is collected through cloud servers and driving vehicles. The collected information is as follows:

[0095] Millimeter-wave radar detected 3 sudden braking events per kilometer;

[0096] Roadside cameras identified five vehicles that did not maintain a safe distance;

[0097] Weather radar echoes predict that precipitation will increase to 4.1 mm / h in the next 30 minutes;

[0098] Generate a standardized feature matrix after spatiotemporal alignment:

[0099]

[0100] Dynamic weight calculation

[0101] The cloud server calculates the final weight using the basic weight and dynamic correction coefficient:

[0102]

[0103] Real-time risk base value calculation:

[0104] =0.322×0.72+0.260×0.88+0.333×0.67+0.085×0.62=0.694.

[0105] Risk prediction increment calculation:

[0106] LSTM input: risk series for the past 60 minutes (sampling interval 5 minutes)

[0107] Capture of spatiotemporal attention mechanism: Traffic flow 5 kilometers upstream increases by 18% / 5 minutes, and cold air from the northwest will reduce visibility to 200 meters.

[0108] Output prediction value: P(t+30min)=0.217P(t+30min)=0.217

[0109] Adjustment coefficient λ = 0.65 (dynamically set according to prediction confidence);

[0110] Comprehensive risk score

[0111] RiskScore=0.694+0.65×0.217=0.835, which is the red alert in the first-level alert.

[0112] The above description is for inspiration. Based on the above description, relevant personnel can make various changes and modifications without departing from the technical concept of this invention. The technical scope of this invention is not limited to the content of the specification, but must be determined according to the scope of the claims.

Claims

1. A dynamic path prediction and control method for an intelligent vehicle collision avoidance device, characterized by: The following steps are involved: Step 1: Road section marking: the driving vehicle connects to the cloud server and marks the road section through the cloud server; Step 2: The cloud server determines the road section information and sends the risk level of the road section to the driving vehicle as a first-level reminder; Step 3: The cloud server collects information about oncoming vehicles on the same road section and makes a judgment based on the oncoming vehicle's movement information to issue a second-level reminder to the driver; The step three comprises the following steps: S31. Collecting vehicle information of incoming vehicles through a cloud server, and searching for information of incoming vehicles of the same brand on the same road section and in a common interconnected network; S32, collecting oncoming vehicle information by the driving vehicle, and transmitting the oncoming vehicle information in front of the driving vehicle to the cloud server through sensors; S33. Determine the route of the oncoming vehicle. The cloud server, based on the collected information about the oncoming vehicle, connects with the vehicle of the same brand and collects the route of the oncoming vehicle of the same brand. If the oncoming vehicle is not of the same brand, the cloud server predicts the path of the oncoming vehicle based on the route information collected by the driving vehicle. S34: Perform a second-level reminder. If the driving vehicle and the oncoming vehicle are on a dangerous route, the cloud server sends a second-level reminder to the driving vehicle. If the driving vehicle and the oncoming vehicle are not on a dangerous route, the cloud server continues to collect oncoming vehicle information. The step 1 comprises the following steps: S11, vehicle positioning: the driving vehicle communicates with the cloud server, which accurately locates the driving vehicle and communicates with the Beidou satellite; S12, road section positioning: the cloud server analyzes the road section where the driving vehicle is located based on the vehicle's location and the navigation software; S13, road section risk analysis: The cloud server performs a risk analysis on the road section based on the data collected by the driving vehicle sensors and the information collected by the navigation software on the road section; S14: Mark the road section. If it is a risky road section, proceed to step 2; if it is not a risky road section, proceed to step S11; In step S13, the formula for analyzing the road section risk is: ； is the real-time risk base value; ,in, , calculate the information entropy of each parameter in real time to reflect the impact of data uncertainty on the weight; is a nonlinear normalization function, The threshold is determined by the quantile of historical data, k is the risk acceleration coefficient, To predict the risk increment, is the adjustment coefficient, LSTM based on spatiotemporal attention mechanism: , the input includes the risk sequence of the previous 60 minutes, the road network topology matrix , time feature vector .

2. The method for dynamic path prediction and control in an intelligent vehicle collision avoidance device according to claim 1, characterized in that: The formula for the road section risk analysis is added with the cross-influence matrix, which is: Where i and j are risk factor numbers, such as i=1 for rainfall intensity and j=2 for emergency braking frequency; k is the time window index, which is a slice every 5 minutes; l is the spatial location code, which is based on the graph node number of the road network topology. The formula for road section risk analysis is as follows: 。 3. The method for dynamic path prediction and control in an intelligent vehicle collision avoidance device according to claim 1, characterized in that: described , The value is adjustable.

4. The method for dynamic path prediction and control in an intelligent vehicle collision avoidance device according to claim 1, characterized in that: The cloud server predicts the route of the oncoming vehicle through a spatiotemporal trajectory prediction model.

5. The method for dynamic path prediction and control in an intelligent vehicle collision avoidance device according to claim 1, characterized in that: The first-level reminder has four reminder levels: green, yellow, orange and red, and the second-level reminder is red.

6. The method for dynamic path prediction and control in an intelligent vehicle collision avoidance device according to claim 1, characterized in that: The driving vehicle is equipped with a 128-line laser radar, a 4K surround-view camera and a millimeter-wave radar.

7. The method for dynamic path prediction and control in an intelligent vehicle collision avoidance device according to claim 5, characterized in that: described The value of is between 0 and 1. At [0,0.4), the first level reminder is green, At [0.4,0.6), the first level reminder is yellow, At [0.6,0.8), the first level reminder is orange. .8, the first-level reminder is red.

Images