Method for identifying auxiliary driving high-speed ramp

Abstract

The invention discloses an auxiliary driving high-speed ramp entrance identification method, which comprises the following steps of S1, data preprocessing: checking and processing navigation data, vehicle driving information, visual perception and radar perception results during vehicle driving; s2, based on the navigation data and the driving information of the vehicle, carrying out fusion positioning on the vehicle and a ramp entrance, and activating an identification algorithm within a certain distance from the vehicle at the ramp entrance and exit; after identification is activated, a visual lane line result sensed based on foresight and a static target result sensed based on radar are subjected to ramp entrance identification and scoring, and an identification result with the highest score is output; s3, repeating the navigation data, the driving information of the vehicle and the visual perception and radar perception results input in the step S1 to obtain a latest positioning and recognition result; s4, carrying out risk assessment on the identified ramp entrance information; according to the method, the problems of ramp positioning failure caused by navigation signal loss, high cost of a high-precision map and long updating period are effectively solved.

Description

Technical Field

[0001] The present invention relates to the field of vehicle assisted driving technology, and in particular to a method for identifying a high-speed ramp entrance by assisted driving. Background Art

[0002] In advanced driver assistance systems (ADAS) and autonomous driving technologies, accurate identification of highway ramps is a core function for safe lane changes, path planning, and navigation. When entering or exiting a highway, vehicles must detect and distinguish between on-ramps (merging in) and off-ramps (merging out) in real time to avoid missing lane changes or entering the wrong lane. The reliability of this function is directly related to driving safety and user experience, and the technical challenges are particularly significant in complex traffic scenarios, such as congestion, at night, or inclement weather.

[0003] Currently, ramp entrance and exit detection in the assisted driving highway NOA process mostly uses visual or high-precision map methods: for example, the method of identifying ramp signs based on visual AI models to plan entrance and exit ramps in advance has good real-time performance and low sensor costs, but the data diversity is insufficient and the robustness to extreme weather scenarios or road repair and change scenarios is poor; another example is the method of identifying ramps based on high-precision maps. This method downloads the HDMap of the ramp area in the cloud through map preloading, which can effectively improve the ramp accuracy and has high planning accuracy. However, the ramp accuracy is affected by positioning accuracy, and the map maintenance cost is high, the road update cycle is long, and the real-time performance is poor.

[0004] The technical background of the highway gate recognition algorithm integrates technologies from multiple fields, including computer vision, sensor fusion, and high-precision positioning. Its development is driving the implementation of Level 3+ assisted driving. In the future, with continuous technological breakthroughs and continuous technical innovation, the algorithm will achieve a qualitative leap in adaptability to complex scenarios, real-time performance, and safety. Summary of the Invention

[0005] The purpose of the present invention is to provide a method for identifying high-speed ramps for assisted driving to solve the problems raised in the above background.

[0006] To achieve the above objectives, the present invention provides the following technical solution: a method for identifying a high-speed ramp entrance for assisted driving, comprising the following steps:

[0007] S1. Data preprocessing: checking and processing the vehicle's navigation data, driving information, visual perception, and radar perception results;

[0008] S2: Based on navigation data and the vehicle's driving information, the vehicle and ramp entrance are integrated and positioned. The recognition algorithm is activated within a certain distance from the vehicle at the ramp entrance or exit. After activation, the ramp entrance is identified and scored based on the visual lane line results perceived by the forward vision and the static target results perceived by the radar. The recognition result with the highest score is output.

[0009] S3, repeat the navigation data, vehicle driving information, visual perception and radar perception results input in step S1 to obtain the latest positioning and recognition results;

[0010] S4. Perform risk assessment on the identified ramp information.

[0011] Preferably, the specific steps of S1 are as follows:

[0012] S11. Determine the validity of the navigation data and extract the ramp endpoints based on the road attributes and other features issued by the navigation data;

[0013] S12. Determine the validity of the vehicle driving information, mainly including whether the input time and parameter range are consistent;

[0014] S13, determining the validity of the lane lines in the visual perception results, mainly including whether the curvature and intercept are consistent;

[0015] S14. Determine the dynamic and static states of the radar perception results, and only retain the static results.

[0016] Preferably, the steps of performing fusion positioning based on navigation data and vehicle driving information in S2 and activating the recognition algorithm within a certain distance from the vehicle at the ramp entrance or exit are as follows:

[0017] S211, constructing RT matrices using navigation data and own vehicle driving data respectively. Based on the navigation signal value, the own vehicle data is used to construct the RT matrix when the navigation signal is poor or lost;

[0018] S212, converting the coordinates of the ramp entrance point in the navigation to the vehicle coordinate system;

[0019] S213. Activate the recognition algorithm at the ramp entrance or exit within a certain distance from the vehicle.

[0020] Preferably, after the recognition in S2 is activated, the ramp entrance will be recognized and scored based on the visual lane line results perceived by the forward vision and the static target results perceived by the radar, and the steps of outputting the recognition result with the highest score are as follows:

[0021] S221, based on the visual lane line result perceived by the forward vision, the ramp entrance is identified based on its curve parameters and virtual-real attributes;

[0022] S222. Based on the radar perception results, extract the static target for edge detection, fit it into curve parameters, and identify the ramp entrance based on the radar edge fitting parameters;

[0023] S223. Score the recognition result and output the ramp recognition result.

[0024] Preferably, in step S223, when scoring the recognition results, the weight distribution of the visual lane line results and the radar perception results is considered, and the weights are dynamically adjusted according to different driving scenarios and sensor reliability.

[0025] Preferably, in step S3, in the process of repeatedly inputting various types of data to obtain the latest positioning and identification results, a data update frequency threshold is set. When the data update frequency is lower than the threshold, it automatically switches to the backup data processing mode to ensure the stability and continuity of the system.

[0026] Preferably, the backup data processing mode includes interpolation using historical data or processing based on model-predicted data.

[0027] Preferably, in step S4, the risk level of entering the ramp is calculated based on factors such as the vehicle's current speed, distance from the ramp, ramp curvature, and traffic flow, and corresponding warning information is provided to the driver based on the risk level.

[0028] The beneficial effects of the present invention are as follows:

[0029] 1. This invention effectively solves the problem of poor robustness of ramp recognition in scenarios such as extreme weather and road changes; it also solves the problems of ramp positioning failure caused by loss of navigation signals, as well as the high cost and long update cycle of high-precision maps. BRIEF DESCRIPTION OF THE DRAWINGS

[0030] Figure 1 This is a flowchart of a technology for identifying high-speed ramps for assisted driving;

[0031] Figure 2 The following is a flowchart of the steps of a method for identifying highway ramps for assisted driving. DETAILED DESCRIPTION

[0032] The technical solution of the present invention is described in detail below with reference to the accompanying drawings and specific embodiments.

[0033] like Figure 1 and Figure 2 As shown, a method for identifying a high-speed ramp entrance by assisted driving of the present invention comprises the following steps:

[0034] S1. Data preprocessing: checking and processing the vehicle's navigation data, driving information, visual perception, and radar perception results.

[0035] Furthermore, the data preprocessing in S1 includes the following steps for checking and processing the navigation data, vehicle driving information, visual perception, and radar perception results during vehicle driving:

[0036] S11. Determine the validity of the navigation data and extract the ramp endpoints based on the road attributes and other features issued by the navigation data;

[0037] The validity mainly includes whether the navigation positioning signal is correct. In the specific implementation process, it is judged by numbers 1 to 5, where 1 is no signal, 2 is a single-point signal, 3 is a differential solution signal, 4 is a floating-point solution signal, and 5 is a fixed solution signal; the results with a signal value greater than or equal to 4 are retained.

[0038] Extract the data with ramp attributes from the navigation data in the order of release. The coordinate points at both ends of the ramp are the ramp endpoints. If the current location is a highway section, the subsequent ramp endpoints are the off-ramp points. If the current location is a non-highway section, the subsequent ramp endpoints are the on-ramp points.

[0039] S12. Determine the validity of the vehicle driving information, mainly including whether the input time and parameter range are consistent;

[0040] The validity of the vehicle driving information is determined, and the results where the time, speed and yaw rate are all zero or Nan abnormal values ​​are eliminated.

[0041] S13, determining the validity of the lane lines in the visual perception results, mainly including whether the curvature and intercept are consistent;

[0042] Determine the validity of lane lines and remove results where both curvature and intercept are zero or abnormal values;

[0043] S14. Determine the dynamic and static states of the radar perception results, and only retain the static results.

[0044] Millimeter-wave radar is used to perceive targets, primarily determining dynamic and static targets based on their longitudinal absolute speed, primarily for dynamic and static objects of a certain size and above the ground. Only dynamic target results are retained. In specific implementation, the speed threshold is relaxed to 2 m / s (7.2 km / h) by default to account for radar accuracy issues. Targets with a longitudinal absolute speed below 2 m / s are considered static targets.

[0045] S2: The vehicle and ramp entrance are integrated and positioned based on navigation data and the vehicle's driving information. The recognition algorithm is activated within a certain distance from the vehicle at the ramp entrance and exit. After recognition activation, the ramp entrance is identified and scored based on the visual lane line results perceived by the forward vision and the static target results perceived by the radar. The recognition result with the highest score is output.

[0046] Furthermore, the steps of performing fusion positioning based on navigation data and vehicle driving information in S2 and activating the recognition algorithm within a certain distance from the vehicle at the ramp entrance and exit are as follows:

[0047] S211, constructing RT matrices using navigation data and own vehicle driving data respectively. Based on the navigation signal value, the own vehicle data is used to construct the RT matrix when the navigation signal is poor or lost;

[0048] The steps for constructing the RT matrix from navigation data are as follows:

[0049] The navigation data is the navigation global coordinate system, and the output positioning information is x, y, z, yaw and signal intensity. The RT matrix constructed at time t is: Where T t For [x t ,y t ,z t ] T , R t Based on yaw t Rotation matrix for angle calculation

[0050] When signalintensity does not meet the signal strength requirement, that is, less than 4, it will be based on the RT at the last valid signal moment. t-1 The RT matrix RT at time t is constructed by combining the vehicle speed and yawrate in the vehicle driving data t The steps are as follows:

[0051] Basic formula for angle: Where t is the current time (the default starting time is set to t-1);

[0052] Distance calculation formula: Where speed t is the velocity at time t, Δt is the time difference between time t and time t-1;

[0053] Coordinate update: Where (x t ,y t ) is the vehicle coordinate at time t (the default height coordinate z is not updated in the plane);

[0054] Transformation Matrix: Where Tt The current position of the vehicle in the navigation global coordinate system [x t ,y t ,z t ] T , R t Based on yaw t Rotation matrix for angle calculation

[0055] S212, converting the coordinates of the ramp entrance point in the navigation to the vehicle coordinate system;

[0056] Conversion formula: Among them, P w (x w ,y w ) is the coordinate of the ramp point in the navigation global coordinate system, P c (x c ,y c ) is the coordinate of the vehicle body coordinate system at the moment of conversion;

[0057] S213, activating the recognition algorithm within a certain distance from the vehicle at the ramp entrance or exit;

[0058] Real-time calculation based on the converted P c (x c ,y c ) The distance from the vehicle after the ramp:

[0059]

[0060] When the distance is less than the distance threshold of 2km, the ramp identification algorithm is performed.

[0061] Furthermore, after the recognition in S2 is activated, the ramp entrance will be recognized and scored based on the visual lane line results perceived by the forward vision and the static target results perceived by the radar. The steps for outputting the recognition result with the highest score are as follows:

[0062] S221, based on the visual lane line result perceived by the forward vision, the ramp entrance is identified based on its curve parameters and virtual-real attributes;

[0063] Forward-view lane line perception is mainly defined as the estimation of the objectively existing or fitted lane boundary (lane line) under the vehicle's current driving state. It includes the lane's virtual and real attributes and curve parameters. The curve expression formula is:

[0064] y(x)=c0+c1×x+c2×x 2 +c3×x 3 ,

[0065] Where: the intercept is c0, unit is m; the slope is atan(c1); the curvature is 2×c2, unit is The curvature change rate is 6×c3, unit

[0066] For the on-ramp situation, calculate whether the left and right curves of the current lane intersect within a certain distance:

[0067] In the specific implementation process, the longitudinal coordinate x is iteratively assigned in the range of 0m-200m with a step size of 1m to solve the horizontal coordinate y of the left lane line l (x) and the lateral coordinate y of the right lane line r The difference of (x) delta_y = y r (x)-y l (x);

[0068] If |delta_y|<0.5m exists during the iteration process, it is considered that the left and right curves of the lane intersect;

[0069] If there is an intersection and the left side of the vehicle's driving lane is a dotted line and the right side is a solid line, then the ramp recognition is successful and subsequent ramp changes can be made;

[0070] For the off-ramp situation, calculate whether the left and right curves of the current lane are far apart within a certain distance:

[0071] In the specific implementation process, first input 0m to assign the longitudinal coordinate x, and solve the lateral coordinate difference between the left and right curves of the lane, delta_0=y r (0)-y l (0) Iterate the vertical coordinate x in the range of 0m-200m with a step length of 1m to solve the difference delta_y=y r (x)-y l (x);

[0072] If |delta_y|>2×|delta_0| exists during the iteration process, it is considered that the left and right curves of the lane are far away from each other.

[0073] If there is an intersection and the left side of the vehicle's driving lane is a dotted line and the right side is a solid line, the output ramp recognition is successful and subsequent ramp changes can be made.

[0074] S222. Based on the radar perception results, extract the static target for edge detection, fit it into curve parameters, and identify the ramp entrance based on the radar edge fitting parameters;

[0075] Convert the radar static target results at multiple moments to the current coordinate system and detect the left and right static edges:

[0076] Combined with the RT conversion matrix calculated by positioning, the radar coordinates at multiple moments are converted to the current coordinate system. In specific implementation, the default multiple moments are 16 sets of input data up to the current moment, that is, the coordinate data of the static target from time t-15, time t-14...time t-1, to the current time t;

[0077] The average yaw angle at multiple moments calculated based on the positioning information is then used to split the radar static target into two sets of points on the left and right sides of the vehicle based on the direction of the average angle.

[0078] In the straight-ahead state, the horizontal coordinates of the left and right groups of points are averaged, and the points where the vehicle reaches the average horizontal coordinate are retained. The point clusters are assigned left and right edge point attributes to obtain the edge line point set.

[0079] In the turning state, based on the driving trajectory, the adjacent points of the left and right groups of point sets are iterated respectively, and the points closest to the driving trajectory are retained. The point clusters are assigned left and right edge point attributes to obtain the edge line point set.

[0080] The virtual road edge line is fitted through the static target clustering results to generate the radar edge curve parameter fitting results:

[0081] Perform curve fitting on the cluster point set to obtain the virtual road centerline fitting result. The fitting formula is y(x)=c0+c1×x+c2×x 2 +c3×x 3 ;

[0082] Ramp identification based on fitting parameters:

[0083] For the on-ramp situation, calculate whether the current right edge curve cuts into the lane within a certain distance:

[0084] Keep the right curve of the fitting results at multiple moments (16 sets of data results up to the current moment) and compare the horizontal coordinate value y of the curve parameters at 200 meters at different moments r (x), if the horizontal axis decreases over time, and at the current time t y r If (x) is less than 0, the ramp identification is successful and subsequent ramp changes can be performed.

[0085] For the off-ramp situation, calculate whether the current right edge curve is away from the lane within a certain distance:

[0086] Keep the right curve of the fitting results at multiple moments (16 sets of data results up to the current moment) and compare the horizontal coordinate value y of the curve parameters at 200 meters at different moments r (x), if the horizontal axis increases with time, and at time t yr If (x) is greater than 10m, the ramp identification is successful and subsequent ramp changes can be made.

[0087] S223. Score the recognition result and output the ramp recognition result.

[0088] The successful result of visual recognition is set to 1.0, and the unsuccessful result is set to 0.0; the successful result of radar recognition is set to 0.5, and the unsuccessful result is set to 0.0; the maximum score result is output, and if all are 0, the recognition failure information is output.

[0089] S3: Repeat the navigation data, vehicle driving information, visual perception and radar perception results input in S1 to obtain the latest positioning and recognition results.

[0090] S4. Perform a risk assessment on the identified ramp entrance information; based on factors such as the vehicle's current speed, distance to the ramp entrance, ramp curvature, and traffic flow, calculate the risk level of entering the ramp entrance, and provide corresponding warning information to the driver based on the risk level.

[0091] The above-described embodiments merely illustrate the implementation methods of the present invention and are not to be construed as limiting the scope of the invention, nor are they to impose any formal limitations on the structure of the present invention. It should be noted that a person skilled in the art may make various changes and improvements without departing from the scope of the present invention, and all such changes and improvements fall within the scope of protection of the present invention.

Claims

1. A method for identifying high-speed ramps for assisted driving, characterized by: The following steps are involved: S1. Data preprocessing: checking and processing the vehicle's navigation data, driving information, visual perception, and radar perception results; S2: Based on navigation data and the vehicle's driving information, the vehicle and ramp entrance are integrated and positioned. The recognition algorithm is activated within a certain distance from the vehicle at the ramp entrance or exit. After activation, the ramp entrance is identified and scored based on the visual lane line results perceived by the forward vision and the static target results perceived by the radar. The recognition result with the highest score is output. S3, repeat the navigation data, vehicle driving information, visual perception and radar perception results input in step S1 to obtain the latest positioning and recognition results; S4. Perform risk assessment on the identified ramp information.

2. The method for identifying a highway ramp entrance for assisted driving according to claim 1, characterized in that: The specific steps of S1 are as follows: S11. Determine the validity of the navigation data and extract the ramp endpoints based on the road attributes and other features issued by the navigation data; S12. Determine the validity of the vehicle driving information, including whether the input time and parameter range are consistent; S13. Determine the validity of the lane lines in the visual perception results, including whether the curvature and intercept are consistent; S14. Determine the dynamic and static states of the radar perception results, and only retain the static results.

3. The method for identifying a highway ramp entrance for assisted driving according to claim 1, characterized in that: In S2, the steps for fusion positioning based on navigation data and vehicle driving information and activating the recognition algorithm within a certain distance from the vehicle at the ramp entrance and exit are as follows: S211, constructing RT matrices using navigation data and own vehicle driving data respectively. Based on the navigation signal value, the own vehicle data is used to construct the RT matrix when the navigation signal is poor or lost; S212, converting the coordinates of the ramp entrance point in the navigation to the vehicle coordinate system; S213. Activate the recognition algorithm at the ramp entrance or exit within a certain distance from the vehicle.

4. The method for identifying a highway ramp entrance for assisted driving according to claim 1, characterized in that: After S2 is activated, the ramp will be identified and scored based on the visual lane line results perceived by the forward vision and the static target results perceived by the radar. The steps to output the recognition result with the highest score are as follows: S221, based on the visual lane line result perceived by the forward vision, the ramp entrance is identified based on its curve parameters and virtual-real attributes; S222. Based on the radar perception results, extract the static target for edge detection, fit it into curve parameters, and identify the ramp entrance based on the radar edge fitting parameters; S223. Score the recognition result and output the ramp recognition result.

5. The method for identifying a highway ramp entrance for assisted driving according to claim 4, characterized in that: In step S223, when scoring the recognition results, the weight distribution of the visual lane line results and the radar perception results is considered, and the weight is dynamically adjusted according to different driving scenarios and sensor reliability.

6. The method for identifying a highway ramp entrance for assisted driving according to claim 1, characterized in that: In step S3, when repeatedly inputting various types of data to obtain the latest positioning and identification results, a data update frequency threshold is set. When the data update frequency is lower than the threshold, it automatically switches to the backup data processing mode to ensure the stability and continuity of the system.

7. The method for identifying a highway ramp entrance for assisted driving according to claim 6, characterized in that: The alternative data processing modes include interpolation using historical data or processing based on model-based predicted data.

8. The method for identifying a highway ramp entrance for assisted driving according to claim 1, characterized in that: In step S4, the risk level of entering the ramp is calculated based on factors such as the vehicle's current speed, distance to the ramp, ramp curvature, and traffic flow, and corresponding warning information is provided to the driver based on the risk level.

Images