Real-time environment construction method fusing perception and navigation map

Abstract

The invention provides a real-time environment construction method fusing perception and a navigation map. The method comprises the following steps: step 1, obtaining fusion perception information and navigation information; 2, obtaining semantic matching information, and dividing a scene into a road section scene and an intersection scene; step 3, respectively performing map construction according to preset construction rules; 4, acquiring navigation information of the intersection scene, and performing curve fitting; 5, calculating the cost value of the topological connection relation of the road information group; 6, calculating the cost values of the center lines of the front and rear lanes with the topological relation, and judging whether the center lines of the two lanes need to be smoothed or not; step 7, when a preset smoothing condition is satisfied, smoothing the distance X of each lane center line before and after cutting off to form a lane center line; and step 8, starting from the lane center line without the precursor relation, iteratively searching the subsequent relation until the searched object has no subsequent relation, and constructing a lane-level guide line.

Description

Technical Field

[0001] The present application relates to the field of autonomous driving technology, and in particular to a method for constructing a real-time environment by integrating perception and navigation maps. Background Art

[0002] The rapid implementation of autonomous driving relies on high-precision map data. Rich in data elements and extremely high in precision, it easily provides global path planning and local, real-time guidance for autonomous driving, effectively acting as a guiding light for the journey. However, high-precision map data also faces challenges. China's road construction is rapidly developing, and road data is updated almost daily. This necessitates that high-precision map data remain fresh to ensure the safety and reliability of autonomous driving. At the same time, the costs of collecting and producing high-precision map data remain high, placing a heavy burden on major autonomous driving companies. While enjoying the convenience of high-precision maps, they also face significant cost pressures. To overcome the bottlenecks posed by high-precision maps for autonomous vehicles, extensive research has been conducted on "mapless" solutions, such as real-time mapping. The challenging task of real-time mapping is to construct accurate environmental information near the vehicle in complex and ever-changing scenarios, such as those with changing lane numbers and complex intersections. In these scenarios, erroneous environmental information can lead to autonomous vehicles entering the wrong lane or experiencing poor navigation. Summary of the Invention

[0003] In view of this, the present invention provides a real-time environment construction method that integrates perception and navigation maps to solve the technical problems in the existing technology, such as the instability and limitations of the visual information of semantic maps, poor adaptability to multi-lane mergers and divisions, intersections, and irregular road sections, and the inability to construct complete road topology information within a local range, such as the inability to cover scenes such as areas without lane lines and intersections.

[0004] The present invention provides a real-time environment construction method for fusion perception and navigation map, the method comprising: step 1, obtaining fusion perception information and navigation information, wherein the navigation information comprises a segmented road information group; step 2, calibrating the navigation information with reference to the fusion perception information, thereby obtaining semantic matching information, and dividing the scene into a road section scene and an intersection scene according to the semantic matching information; step 3, dividing the road section scene into a basic road section, a road section with a changed number of lanes, and an irregular road section, and constructing lane lines respectively according to preset construction rules to form a road information group; step 4, obtaining navigation information of the intersection scene, performing curve fitting according to the entry and exit points and the waypoints in the navigation information, and if the road information to which the entry and exit points of the fitted curve belong is the same, the fitted curve is classified into one road information group; step 5, designing the following cost equation according to the road information groups constructed by the road sections and intersections, and calculating the cost value of the topological connection relationship of the road information group, cost (topo) =δ dir * direction(lane)+δ dis * distance(lane)+δ angle * angle(lane), where cost (topo) represents the cost of two road information groups under this evaluation system, δ dir represents the direction weight of the lane centerline lane, and the direction cost direction(lane) represents the calculated value in the local coordinate system of the lane centerline lane. dis is the weight of the distance cost. The distance cost distance(lane) represents the minimum distance calculated value in the local coordinate system of the center lines of the two lanes. angle Represents the weight of the angle deviation, and the angle cost angle(lane) represents the calculated value of the angle deviation of the center lines of the two lanes in the local coordinate system; Step 6, according to the following cost equation, calculate the cost value of the center lines of the front and rear lanes with topological relationship, and determine whether the center lines of the two lanes need to be smoothed. (smooth) = δ dis * distance(lane)+δ angle * angle(lane), cost (smooth) represents the cost value of the center lines of the two lanes under the evaluation system, and the distance cost distance(lane) represents the minimum distance calculated value of the center lines of the two lanes in the local coordinate system. angle Indicates the weight of the angle deviation, and the angle cost angle(lane) represents the angle deviation calculated value in the local coordinate system of the center line lane of the two lanes; Step 7, when cost (smooth)When the value meets the preset smoothing conditions, the X distance between the front and rear lane centerlines of each cutoff is smoothed according to the preset smoothing range parameters to form a lane centerline. In step 8, starting from the lane centerline without a predecessor relationship, its successor relationship is iteratively searched until the search object has no successor relationship. All the searched lane centerlines are constructed into a complete lane-level guide line, thus completing the environment construction.

[0005] Furthermore, the fused perception information includes lane line information, road boundary information, and stop line information obtained by sensors; the navigation information includes lane line information, road boundary information, stop line information, and intersection information collected by high-precision maps; and the road information group includes segmented lane lines, road boundary information, and semantic information related thereto.

[0006] Furthermore, the construction rules include: on the basic road section, based on the two adjacent lane lines on the left and right, the lane center line is constructed by geometric point selection or translation. The lane center line point calculation formula is: Centerlane pos =(leftlane pos +rightlane pos )*0.5, where Centerlane pos is the lane centerline, leftlane pos For the left lane line, rightlane pos For the right lane line; on the road section with varying lane numbers, if the width change of the left and right adjacent lane lines does not exceed the distance threshold, the lane center line is generated using the basic road section construction method; if the width change of the left and right adjacent lane lines exceeds the distance threshold, the lane center line is generated according to the calculation formula , Generate lane centerline, where Centerlane pos is the lane centerline, leftlane pos For the left lane line, rightlane pos For the right lane line, leftlane length is the length of the left lane line, rightlane length is the length of the right lane line, and d is the distance from the current calculation point to the starting point of the lane line on the same side. On irregular road sections, if the traffic mode is left turn or straight line, the left lane line is translated to construct the lane centerline. If the traffic mode is right turn, the right lane line is translated to construct the lane centerline.

[0007] Furthermore, the left lane line is translated to construct the lane center line by taking the left lane linear point as the reference and shifting it to the right by a distance of L1 to form the lane center line; the right lane line is translated to construct the lane center line by taking the right lane linear point as the reference and shifting it to the left by a distance of d2-L1, where L1 is the distance from the exit point to the left lane line, and d2 is the distance from the left lane line to the end point of the right lane line.

[0008] Furthermore, the direction cost, distance cost and angle cost in the topological connection relationship cost formula of the road information group are normalized.

[0009] Furthermore, the distance cost and angle cost in the cost formula of the center lines of the front and rear lanes are normalized.

[0010] Furthermore, the center lines of the front and rear lanes are smoothed using the OSQP open source library.

[0011] Furthermore, step 8 also includes: combining the fused perception information and the needs of the downstream modules, giving the constructed lane guide lines relevant attributes so that they can be directly used in the planning module, thereby avoiding the downstream modules from incurring additional time and resource losses due to operations such as data construction.

[0012] Furthermore, the relevant attributes include the distance from the center point of the lane line to the left and right lane lines, the distance from the center point of the lane line to the road boundary, whether there is a stop line at the center point of the lane, and whether there is a merging or diverging scenario.

[0013] The present invention provides a real-time environment construction method that integrates perception and navigation maps. By integrating perception and navigation information to complete real-time mapping, the application significantly reduces the cost of autonomous driving while ensuring performance and stability. The application also proposes a scenario-based partitioning method. This method effectively constructs local environments in scenarios with varying lane numbers, irregular road sections, and irregular intersections, improving the scene richness and robustness of real-time autonomous driving mapping, thereby indirectly enhancing the comfort and stability of autonomous driving. BRIEF DESCRIPTION OF THE DRAWINGS

[0014] Figure 1 This is a flow chart of a method for constructing a real-time environment by integrating perception and navigation maps provided by the present invention; Figure 2 This is a schematic diagram of a road information group provided by the present invention; Figure 3 It is a schematic diagram of the single lane structure provided by the present invention; Figure 4 This is a schematic diagram of lane lines constructed on a road section with a changed number of lanes provided by the present invention; Figure 5This is a schematic diagram of constructing lane lines on irregular road sections provided by the present invention; Figure 6 It is a schematic diagram of the smoothing and connection relationship structure provided by the present invention. DETAILED DESCRIPTION

[0015] The following will clearly and completely describe the technical solutions in the embodiments of the present invention in conjunction with the accompanying drawings. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without making creative efforts are within the scope of protection of the present invention.

[0016] Example 1: The present invention provides a real-time environment construction method integrating perception and navigation map. Figure 1 As shown, the method includes the following steps.

[0017] Step 1: Acquire fused perception information and navigation information, where the navigation information includes segmented road information groups; Step 2: Using the fused perception information as a reference, the navigation information is calibrated to obtain semantic matching information. Based on the semantic matching information, the scene is divided into a road section scene and an intersection scene. Step 3: Divide the road scene into basic sections, sections with varying lane numbers, and irregular sections, and construct maps for each according to preset construction rules; Step 4: Obtain navigation information for the intersection scene, perform curve fitting based on the entry and exit points and waypoints in the navigation information, and if the road information of the entry and exit of the fitted curve is the same, then the fitted curve is classified into one road information group; Step 5: Based on the road information group constructed by road sections and intersections, the following cost equation is designed to calculate the cost value of the topological connection relationship of the road information group: cost (topo) = δ dir * direction(lane)+δ dis * distance(lane)+δ angle * angle(lane), Among them, cost (topo) represents the cost of two road information groups under this evaluation system, δ dir represents the direction weight of the lane centerline lane, and the direction cost direction(lane) represents the calculated value in the local coordinate system of the lane centerline lane. disis the weight of the distance cost. The distance cost distance(lane) represents the minimum distance calculated value in the local coordinate system of the center lines of the two lanes. angle The angle cost angle(lane) represents the angle deviation calculated in the local coordinate system of the center lines of the two lanes. Step 6: Calculate the cost value of the center lines of the front and rear lanes with topological relationship according to the following cost equation to determine whether the center lines of the two lanes need to be smoothed. cost (smooth) = δ dis * distance(lane)+δ angle * angle(lane), cost (smooth) represents the cost value of the center lines of the two lanes under the evaluation system, and the distance cost distance(lane) represents the minimum distance calculated value of the center lines of the two lanes in the local coordinate system. angle The angle cost angle(lane) represents the angle deviation calculated in the local coordinate system of the center lines of the two lanes. Step 7, when cost (smooth) When the value meets the preset smoothing conditions, the X distance between the lane center lines before and after each cutoff is smoothed according to the preset smoothing range parameters to form a lane center line; Step 8: Starting from the lane centerline without a predecessor relationship, iteratively search its successor relationship until the search object has no successor relationship. All the searched lane centerlines are constructed into a complete lane-level guide line, thus completing the environment construction.

[0018] The present invention provides a real-time environment construction method that integrates perception and navigation maps. By integrating perception and navigation information to complete real-time mapping, the application significantly reduces the cost of autonomous driving while ensuring performance and stability. The application also proposes a scenario-based partitioning method. This method effectively constructs local environments in scenarios with varying lane numbers, irregular road sections, and irregular intersections, improving the scene richness and robustness of real-time autonomous driving mapping, thereby indirectly enhancing the comfort and stability of autonomous driving.

[0019] Example 2: The present invention provides a real-time environment construction method integrating perception and navigation map. Figure 1 As shown, the method includes the following steps.

[0020] Step 1: Acquire fused perception information and navigation information, where the navigation information includes segmented road information groups; The fused perception information includes lane line information, road boundary information, and stop line information obtained through sensors (lidar, cameras, radars, etc.). The navigation information includes lane line information, road boundary information, stop line information, and intersection information collected by high-precision maps. High-precision maps, also known as navigation maps, are electronic maps designed specifically for vehicle navigation, autonomous driving, or other location services. They use geographic information and road data to provide users with optimal path planning, real-time road condition updates, and location guidance from the starting point to the end point. It has lane lines, road boundaries, stop lines, intersections, and rich POIs and other element information. Different elements also have correlations. A road is usually segmented into road information groups according to certain rules. The road information group includes segmented lane lines, road boundary information, and related semantic information. It includes lane line point sets and their semantic information, road boundary line point sets and their semantic information, and stop line point sets and their semantic information. The lane lines and road boundaries in the road information group are point information, and the semantic information is the spatial relationship between elements, such as Figure 2 shown.

[0021] Step 2: Using the fused perception information as a reference, the navigation information is calibrated to obtain semantic matching information. Based on the semantic matching information, the scene is divided into a road section scene and an intersection scene. Navigation information provides segmented road information groups. Using fusion perception as a reference, the navigation information is dynamically calibrated. The calibrated information is collectively referred to as semantic matching information. Based on this semantic matching information, the scene is divided into road sections (with lane lines) and intersection scenes (without lane lines).

[0022] Step 3: Divide the road scene into basic sections, sections with varying lane numbers, and irregular sections, and construct maps for each according to preset construction rules; Road scenes are divided into three types: basic sections, sections with varying lane numbers, and irregular sections, and maps are constructed according to the following rules.

[0023] The first type is the basic road section, which is a straight or curved road section. The characteristics of the basic road section are that the lane trend is uniform and the width does not change significantly on the basic road section. On the basic road section, based on the two adjacent lane lines, the lane centerline is constructed by geometric point selection (taking the midpoint of the left and right lane lines) or translation (translating the left lane line to the right). The lane centerline calculation formula is: Centerlane pos =( leftlane pos +rightlane pos )*0.5, where Centerlane pos is the lane centerline, leftlane posFor the left lane line, rightlane pos is the right lane line, and the single lane structure diagram is as follows Figure 3 shown.

[0024] The second type is a road section with a changed number of lanes, where the width of the adjacent lane lines on the left and right sides is less than the preset threshold. On a road section with a changed number of lanes, if the width change of the adjacent lane lines on the left and right sides does not exceed the distance threshold, the lane centerline is generated using the basic road section construction method. If the width change of the adjacent lane lines on the left and right sides exceeds the distance threshold, the lane centerline is generated according to the calculation formula. , Generate lane centerline, where Centerlane pos is the lane centerline, leftlane pos For the left lane line, rightlane pos For the right lane line, leftlane length is the length of the left lane line, rightlane length is the length of the right lane line, and d is the distance from the current calculation point to the starting point of the lane line on the same side. Specifically, if the width of the adjacent lane lines on the left and right sides changes beyond the distance threshold, it is necessary to determine whether the lane changes to the right or the left. If it changes to the left, the constructed lane line is no longer adjacent to the left and right, but spans a lane, such as Figure 4 As shown, the calculation formula is ; If it changes to the right, the calculation formula is .

[0025] The third type is irregular road sections, that is, road sections with uneven width changes near intersections. The points of the fitting curve in the intersection are fixed (entry and exit points, passing points, see intersection scene construction for detailed description), which are derived from semantic matching information. This point is more inclined to the actual vehicle operation route; the lane centerline of the road section is mostly generated in the center of the road. Due to the irregularity of the road section, the overall shape of the road section is similar to a trapezoid. The position of the generated centerline will deviate from the point of the intersection and cannot be connected, and a better curve cannot be fitted through smoothing. In order to solve the above problems, on irregular road sections, refer to the actual vehicle operation rules and adjust the generation method of the road section in real time according to the traffic mode of the intersection: if the traffic mode is left turn or straight line, the left lane line is translated to construct the lane centerline; if the traffic mode is right turn, the right lane line is translated to construct the lane centerline. Such as Figure 5 As shown, when the end of the lane line group is a continuous intersection (entrance), if the road direction is straight (the vehicle turns left or goes straight), the left lane line is translated to construct the lane center line. Taking the left lane linear point as the reference, it is translated to the right by a distance of L1 to form the lane center line. The specific calculation formula is Centerlanepos = leftlane pos +L1; If the road turns right (the vehicle turns right), the right lane line is translated to construct the lane centerline. Based on the right lane line point, it is translated to the left by a distance of d2-L1, where L1 is the distance from the exit point to the left lane line, and d2 is the distance from the left lane line to the end point of the right lane line.

[0026] Step 4: Obtain navigation information for the intersection scene, perform curve fitting based on the entry and exit points and waypoints in the navigation information, and if the road information of the entry and exit of the fitted curve is the same, then the fitted curve is classified into one road information group; Step 5: Based on the road information group constructed by road sections and intersections, the following cost equation is designed to calculate the cost value of the topological connection relationship of the road information group: cost (topo) = δ dir * direction(lane)+δ dis * distance(lane)+δ angle * angle(lane), Among them, cost (topo) represents the cost of two road information groups under this evaluation system, δ dir represents the direction weight of the lane centerline lane, and the direction cost direction(lane) represents the calculated value in the local coordinate system of the lane centerline lane. dis is the weight of the distance cost. The distance cost distance(lane) represents the minimum distance calculated value in the local coordinate system of the center lines of the two lanes. angle The angle cost angle(lane) represents the angle deviation calculated in the local coordinate system of the center lines of the two lanes. Step 6: Calculate the cost value of the center lines of the front and rear lanes with topological relationship according to the following cost equation to determine whether the center lines of the two lanes need to be smoothed. cost (smooth) = δ dis * distance(lane)+δ angle * angle(lane), cost (smooth) represents the cost value of the center lines of the two lanes under the evaluation system, and the distance cost distance(lane) represents the minimum distance calculated value of the center lines of the two lanes in the local coordinate system. angleThe angle cost angle(lane) represents the angle deviation calculated in the local coordinate system of the center lines of the two lanes. Step 7, when cost (smooth) When the value meets the preset smoothing conditions, the X distance between the lane center lines before and after each cutoff is smoothed according to the preset smoothing range parameters to form a lane center line; Step 8: Starting from the lane centerline without a predecessor relationship, iteratively search its successor relationship until the search object has no successor relationship. All the searched lane centerlines are constructed into a complete lane-level guide line, thus completing the environment construction.

[0027] The present invention provides a real-time environment construction method that integrates perception and navigation maps. By integrating perception and navigation information to complete real-time mapping, the application significantly reduces the cost of autonomous driving while ensuring performance and stability. The application also proposes a scenario-based partitioning method. This method effectively constructs local environments in scenarios with varying lane numbers, irregular road sections, and irregular intersections, improving the scene richness and robustness of real-time autonomous driving mapping, thereby indirectly enhancing the comfort and stability of autonomous driving.

[0028] Example 3: The present invention provides a real-time environment construction method integrating perception and navigation map. Figure 1 As shown, the method includes the following steps.

[0029] Step 1: Acquire fused perception information and navigation information, where the navigation information includes segmented road information groups; Step 2: Using the fused perception information as a reference, the navigation information is calibrated to obtain semantic matching information. Based on the semantic matching information, the scene is divided into a road section scene and an intersection scene. Step 3: Divide the road scene into basic sections, sections with varying lane numbers, and irregular sections, and construct maps for each according to preset construction rules; Step 4: Obtain navigation information for the intersection scene, perform curve fitting based on the entry and exit points and waypoints in the navigation information, and if the road information of the entry and exit of the fitted curve is the same, then the fitted curve is classified into one road information group; Step 5: Based on the road information group constructed by road sections and intersections, the following cost equation is designed to calculate the cost value of the topological connection relationship of the road information group: cost (topo) = δ dir * direction(lane)+δ dis * distance(lane)+δ angle* angle(lane), Among them, cost (topo) represents the cost of two road information groups under this evaluation system, δ dir represents the direction weight of the lane centerline lane, and the direction cost direction(lane) represents the calculated value in the local coordinate system of the lane centerline lane. dis is the weight of the distance cost. The distance cost distance(lane) represents the minimum distance calculated value in the local coordinate system of the center lines of the two lanes. angle The angle cost angle(lane) represents the angle deviation calculated in the local coordinate system of the center lines of the two lanes. The direction cost, distance cost and angle cost in the topological connection relationship cost formula of the road information group are normalized.

[0030] Step 6: Calculate the cost value of the center lines of the front and rear lanes with topological relationship according to the following cost equation to determine whether the center lines of the two lanes need to be smoothed. cost (smooth) = δ dis * distance(lane)+δ angle * angle(lane), cost (smooth) represents the cost value of the center lines of the two lanes under the evaluation system, and the distance cost distance(lane) represents the minimum distance calculated value of the center lines of the two lanes in the local coordinate system. angle The angle cost angle(lane) represents the angle deviation calculated in the local coordinate system of the center lines of the two lanes. The distance cost and angle cost in the cost formula of the center lines of the front and rear lanes are normalized.

[0031] Step 7, when cost (smooth) When the value meets the preset smoothing conditions, the X distance between the lane center lines before and after each cutoff is smoothed according to the preset smoothing range parameters to form a lane center line; If the smoothing conditions are met, the X distance between the center lines of the front and rear lanes is smoothed according to the smoothing range parameters to avoid distortion of the shape points due to excessive smoothing. The center lines of the front and rear lanes are smoothed using the OSQP open source library. The schematic diagram of the smoothing and connection relationship is shown in the figure. Figure 6As shown in the figure, in addition to smoothing, the OSQP open-source library can also be used for curve fitting, as described above. OSQP (Operator Splitting Quadratic Programming Solver) is a numerical optimization package for solving convex quadratic programming problems. Based on the regularized Newton method, it effectively handles quadratic programming problems with both equality and inequality constraints.

[0032] Step 8: Starting from the lane centerline without a predecessor relationship, iteratively search its successor relationship until the search object has no successor relationship. All the searched lane centerlines are constructed into a complete lane-level guide line, thus completing the environment construction.

[0033] Step 8 also involves assigning relevant attributes to the constructed lane guide lines, combining the fused perception information with the needs of downstream modules. This allows them to be directly used in the planning module, avoiding the additional time and resource consumption of downstream modules due to operations such as data construction. These attributes include the distance from the lane center point to the left and right lane lines, the distance from the lane center point to the road boundary, whether there is a stop line at the lane center point, and whether there is a merge / diverge scenario.

[0034] The present invention provides a real-time environment construction method that integrates perception and navigation maps. By integrating perception and navigation information to complete real-time mapping, the application significantly reduces the cost of autonomous driving while ensuring performance and stability. The application also proposes a scenario-based partitioning method. This method effectively constructs local environments in scenarios with varying lane numbers, irregular road sections, and irregular intersections, improving the scene richness and robustness of real-time autonomous driving mapping, thereby indirectly enhancing the comfort and stability of autonomous driving.

[0035] In summary, the embodiments of the present invention provide a real-time environment construction method that integrates perception and navigation maps. This technical solution is a dynamic map construction technology for autonomous vehicles. By integrating data from on-board sensors such as lidar, cameras, and radar, it captures and updates detailed information about the vehicle's surroundings in real time, including road topology, lane centerlines, and traffic signs. This allows for the construction of lane-level maps, achieving capabilities similar to lane-level navigation using high-precision maps. This eliminates the need for HD maps and enables accurate and stable lane construction in "no map" situations, ensuring safe driving.

[0036] The above description is only a preferred embodiment of the present invention and is not intended to limit the present invention. Any modifications, equivalent substitutions, improvements, etc. made within the spirit and principles of the present invention should be included in the scope of protection of the present invention.

Claims

1. A real-time environment construction method integrating perception and navigation map, characterized in that: The method comprises: Step 1: Acquire fused perception information and navigation information, where the navigation information includes segmented road information groups; Step 2: Using the fused perception information as a reference, the navigation information is calibrated to obtain semantic matching information. Based on the semantic matching information, the scene is divided into a road section scene and an intersection scene. Step 3: Divide the road scene into basic road sections, road sections with varying lane numbers, and irregular road sections, and construct lane lines for each section according to preset construction rules to form a road information group; Step 4: Obtain navigation information for the intersection scene, perform curve fitting based on the entry and exit points and waypoints in the navigation information, and if the road information of the entry and exit of the fitted curve is the same, then the fitted curve is grouped into one road information group; Step 5: Based on the road information group constructed by road sections and intersections, the following cost equation is designed to calculate the cost value of the topological connection relationship of the road information group: cost (topo) = δ dir * direction(lane)+ δ dis * distance(lane)+ δ angle * angle(lane), Among them, cost (topo) represents the cost of two road information groups under this evaluation system, δ dir represents the direction weight of the lane centerline lane, and the direction cost direction(lane) represents the calculated value in the local coordinate system of the lane centerline lane. dis is the weight of the distance cost. The distance cost distance(lane) represents the minimum distance calculated value in the local coordinate system of the center lines of the two lanes. angle The angle cost angle(lane) represents the angle deviation calculated in the local coordinate system of the center lines of the two lanes. Step 6: Calculate the cost value of the center lines of the front and rear lanes with topological relationship according to the following cost equation to determine whether the center lines of the two lanes need to be smoothed. cost (smooth) = δ dis * distance(lane)+ δ angle * angle(lane), cost (smooth) represents the cost value of the center lines of the two lanes under the evaluation system, and the distance cost distance(lane) represents the minimum distance calculated value of the center lines of the two lanes in the local coordinate system. angle The angle cost angle(lane) represents the angle deviation calculated in the local coordinate system of the center lines of the two lanes. Step 7, when cost (smooth) When the value meets the preset smoothing conditions, the X distance between the lane center lines before and after each cutoff is smoothed according to the preset smoothing range parameters to form a lane center line; Step 8: Starting from the lane centerline without a predecessor relationship, iteratively search its successor relationship until the search object has no successor relationship. All the searched lane centerlines are constructed into a complete lane-level guide line, thus completing the environment construction.

2. The method for constructing a real-time environment integrating perception and navigation map according to claim 1, characterized in that: The fused perception information includes lane line information, road boundary information, and stop line information obtained by sensors; the navigation information includes lane line information, road boundary information, stop line information, and intersection information collected by high-precision maps; and the road information group includes segmented lane lines, road boundary information, and semantic information related thereto.

3. The method for constructing a real-time environment integrating perception and navigation map according to claim 1, characterized in that: The construction rules include: On the basic road section, based on the two adjacent lane lines on the left and right, the lane centerline is constructed by geometric point selection or translation. The lane centerline point calculation formula is: Centerlane pos =(leftlane pos +rightlane pos )*0.5, where Centerlane pos is the lane centerline, leftlane pos For the left lane line, rightlane pos For the right lane line; On a road section with a changed number of lanes, if the width change of the left and right adjacent lane lines does not exceed the distance threshold, the lane centerline is generated using the basic road section construction method. If the width change of the left and right adjacent lane lines exceeds the distance threshold, the lane centerline is generated according to the calculation formula ， Generate lane centerline, where Centerlane pos is the lane centerline, leftlane pos For the left lane line, rightlane pos For the right lane line, leftlane length is the length of the left lane line, rightlane length is the length of the right lane line, and d is the distance from the current calculation point to the starting point of the lane line on the same side; On irregular road sections, if the mode of travel is a left turn or a straight line, the left lane line is translated to construct the lane center line. If the mode of travel is a right turn, the right lane line is translated to construct the lane center line.

4. The method for constructing a real-time environment integrating perception and navigation map according to claim 3, characterized in that: The left lane line is translated to construct the lane center line by taking the left lane linear point as the reference and shifting it to the right by a distance of L1 to form the lane center line; the right lane line is translated to construct the lane center line by taking the right lane linear point as the reference and shifting it to the left by a distance of d2-L1, where L1 is the distance from the exit point to the left lane line, and d2 is the distance from the left lane line to the end point of the right lane line.

5. The method for constructing a real-time environment integrating perception and navigation map according to claim 1, characterized in that: The direction cost, distance cost and angle cost in the topological connection relationship cost formula of the road information group are normalized.

6. The method for constructing a real-time environment integrating perception and navigation map according to claim 1, characterized in that: The distance cost and angle cost in the cost formula of the center lines of the front and rear lanes are normalized.

7. The method for constructing a real-time environment integrating perception and navigation map according to claim 1, characterized in that: The center lines of the front and rear lanes are smoothed using the OSQP open source library.

8. The method for constructing a real-time environment integrating perception and navigation map according to claim 1, characterized in that: The step 8 also includes: combining the fused perception information with the needs of downstream modules, giving the constructed lane guide lines relevant attributes so that they can be directly used in the planning module, thereby avoiding the downstream modules from incurring additional time and resource loss due to operations such as data construction.

9. The method for constructing a real-time environment integrating perception and navigation map according to claim 8, characterized in that: The relevant attributes include the distance from the center point of the lane line to the left and right lane lines, the distance from the center point of the lane line to the road boundary, whether there is a stop line at the center point of the lane, and whether there is a merging or diverging scenario.

Images