A lane-link and road shape information based driving system and driving method
The lane link and road shape information-based deflection driving system addresses inaccuracies in conventional systems by generating control parameters for vehicle deflection, ensuring stable and comfortable driving through accurate consideration of surrounding vehicles and road conditions.
Patent Information
- Application Number
- KR1020200186460
- Authority / Receiving Office
- KR · KR
- Patent Type
- Patents
- Current Assignee / Owner
- Filing Date
- 2020-12-29
- Publication Date
- 2026-11-10
- Estimated Expiration
- 2040-12-29
AI Technical Summary
Conventional vehicle deflection driving systems fail to accurately consider self-vehicle position recognition, risk of surrounding vehicles, driving tendencies, and road curvature, leading to discontinuities in bias control and unnatural vehicle behavior, especially in complex driving scenarios like lane changes and intersections.
A lane link and road shape information-based deflection driving system that generates control parameters using virtual lines and control values to manage vehicle deflection by integrating map information, surrounding vehicle positions, and risk levels, while accounting for driver tendencies and road shape.
Ensures stable and comfortable vehicle deflection control in complex driving conditions, maintaining safe distances and smooth lane changes by considering vehicle type, location accuracy, and driver tendencies, even on curved roads.
Smart Images

Figure 112020142855359-PAT00004_ABST
Abstract
Description
Technology Field
[0001] The present invention relates to a deflection driving system, and in particular, to a lane link that controls the deflection driving of a vehicle by considering the accuracy of self-vehicle position recognition, the risk of surrounding vehicles, driving tendencies, the curvature of the road, and the shape of the road, and to a road shape information-based deflection driving system and a deflection driving method. Background Technology
[0002] For automatic driving of a vehicle, automatic control of acceleration and deceleration, and steering control technology centered on the driving path are required. According to Korean published patent No. 10-2019-0123663 disclosed on November 1, 2019, a driving path providing device including a sensor (110), a steering device (120), an input device (130), and a control circuit (140) is used to drive along the path of a driving lane on one hand and change lanes on the other.
[0003] The sensor (110) acquires information about an external object, and the input device (130) receives a lane change command from the driver of the vehicle. The control circuit (140) controls the vehicle to drive along a deflected path within the driving lane based on the information acquired from the sensor (110) and the operation of the steering device (120), and when a lane change command is received by the input device (130) while driving along the deflected path within the driving lane, the control circuit controls the vehicle to drive along the deflected path within the target lane of the lane change after completing the lane change.
[0004] Conventional bias driving systems controlled vehicle driving by determining bias values based on the safe distance between the vehicle and surrounding vehicles as a parameter. Consequently, when determining the degree of bias in various situations—such as normal driving, lane changes, and intersections—discontinuities in the bias values occurred, resulting in unnatural vehicle bias control. Furthermore, the complexity of the bias value determination logic increased due to the need to involve numerous branching statements, and there was a disadvantage in that factors such as the accuracy of the vehicle's position recognition, the risk level of surrounding vehicles, and the driver's driving style could not be reflected in the bias values.
[0005] In particular, using the degree of lane encroachment as a criterion for determining the bias value causes a problem where the vehicle fails to react in advance, even when approach by other vehicles is anticipated in sections where the lane width narrows. The problem to be solved
[0006] The technical problem that the present invention aims to solve is to propose a lane link and a road shape information-based deflection driving system that controls the deflection driving of a vehicle by considering the accuracy of self-vehicle position recognition, the risk level of surrounding vehicles, driving tendencies, the curvature of the road, and the shape of the road.
[0007] Another technical problem that the present invention aims to solve is to propose a lane link and a road shape information-based deflection driving method that controls the deflection driving of a vehicle by considering the accuracy of self-vehicle position recognition, the risk level of surrounding vehicles, driving tendencies, the curvature of the road, and the shape of the road. means of solving the problem
[0008] A lane link and a road shape information-based deflection driving system according to the present invention for achieving the above technical problem comprises: a control parameter generation unit that generates control parameters using the virtual line, and a control unit that generates control values for a vehicle corresponding to the control parameters received from the control parameter generation unit, and controls the deflection of the vehicle by generating control values for the vehicle that generate control values for the vehicle that control the vehicle's deflection by using fused object information, which is map information including the position and speed of surrounding vehicles, and extracts an object that causes the deflection of the vehicle, and generates a virtual line of the extracted object by reflecting at least one of the risk level of surrounding vehicles, position recognition accuracy, driving tendency of the driver of the vehicle, road curvature, and road shape.
[0009] A lane link and road shape information-based bias driving method according to the present invention for achieving the above other technical objectives comprises: a step of extracting a bias target object based on a bias section; a step of extracting a virtual line by reflecting a predetermined condition to a reference point to be biased and the extracted bias target object; a step of grouping objects corresponding to a predetermined standard among bias objects close to the self vehicle; a step of calculating a bias offset for each grouped object cluster; a step of generating a bias path, which is a path on the lane link reflecting the calculated offset; a step of generating a speed profile of the self vehicle moving along the bias path; and a step of generating control parameters reflecting the bias path and the speed profile.
[0010] The technical problems to be solved by the present invention are not limited to those mentioned above, and other technical problems not mentioned will be clearly understood by those skilled in the art to which the present invention belongs from the description below. Effects of the invention
[0011] The lane link and road shape information-based deflection driving system and lane link and road shape information-based deflection driving method according to the present invention as described above have the advantage of being able to simply obtain a deflection value even for a vehicle driving in a complex manner in an adjacent lane, and to obtain a reliable deflection value in the same way even on curved roads rather than straight roads, and to ensure driving stability and ride comfort by passing through the actual vehicle outlines while considering vehicle type, location recognition accuracy, and driver tendencies, and to be applicable in sections where lane width changes, such as merging lanes, intersections, and toll gates, by reflecting the shape of the road.
[0012] The effects obtainable from the present invention are not limited to those mentioned above, and other unmentioned effects will be clearly understood by those skilled in the art from the description below. Brief explanation of the drawing
[0013] FIG. 1 is an embodiment of a lane link and a road shape information-based deflection driving system according to the present invention. Figure 2 is an example of the configuration of a control parameter generation unit. Figure 3 illustrates the operation of the bias target object extraction module. Figure 4 illustrates the operation of the object virtual line extraction module. Figure 5 is an example of a method for determining a virtual line and a center line when there are vehicles on both sides of the vehicle. Figure 6 illustrates a conventional method for determining a virtual line and a center line when a vehicle exists on only one side of the vehicle. FIG. 7 illustrates the method proposed in the present invention for determining a virtual line and a center line when a vehicle exists on only one side of the vehicle. Figure 8 illustrates a method for reflecting risk levels when setting virtual lines in the object virtual line extraction module. Figure 9 illustrates a method for reflecting position recognition accuracy when setting virtual lines in the object virtual line extraction module. Figure 10 illustrates a method for reflecting the driving tendencies of a vehicle driver when setting virtual lines in an object virtual line extraction module. Figure 11 illustrates a method for reflecting the driving tendencies of the vehicle driver when determining the lane width in the object virtual line extraction module. Figure 12 illustrates a method for reflecting the curvature of a road when setting a virtual line in the object virtual line extraction module. Figure 13 illustrates a method for reflecting the shape of a road when setting a virtual line in the object virtual line extraction module. Figure 14 is an example of a method for determining the final virtual line. FIG. 15 illustrates an example of clustering performed in a biased object clustering module. FIG. 16 illustrates another embodiment of clustering performed in the biased object clustering module. FIG. 17 illustrates another example of clustering performed in the biased object clustering module. Figure 18 illustrates the bias path generated by the bias path generation module. FIG. 19 is an example of a lane link and a road shape information-based deflection driving method according to the present invention. Figure 20 illustrates an example of bias depending on the presence or absence of a vehicle in the opposite lane while driving in the outermost lane. FIG. 21 illustrates an example of the application of the present invention in a situation where dangerous vehicles and ordinary vehicles are mixed. Specific details for implementing the invention
[0014] In order to fully understand the present invention, the operational advantages of the present invention, and the objectives achieved by the implementation of the present invention, reference must be made to the accompanying drawings describing exemplary embodiments of the present invention and the contents described in the accompanying drawings.
[0015] The present invention will be described in detail below by explaining preferred embodiments of the invention with reference to the attached drawings. Identical reference numerals in each drawing indicate identical components.
[0016] FIG. 1 is an embodiment of a lane link and a road shape information-based deflection driving system according to the present invention.
[0017] Referring to FIG. 1, the lane link and road shape information-based deflection driving system (200, hereinafter deflection driving system) according to the present invention includes a perception sensor unit (210), a precision map transmission unit (220), a GPS receiver unit (230), a communication unit (240), a deflection object information generation unit (250), a control parameter generation unit (260), and a control unit (270).
[0018] The perception sensor unit (210) includes a lidar (211), a camera (212), and a radar (213), and generates perception information by collecting information about the driving road, surrounding information of the driving road, and surrounding vehicles. The precision map transmission unit (220) provides a precision map of the surroundings of the vehicle equipped with the deflection driving system (200).
[0019] The GPS receiver (230) receives a signal sent from a GPS satellite (not shown) or uses the received signal to calculate the current location of the vehicle, i.e., the vehicle equipped with the bias driving system (200). The communication unit (240) is a means for transmitting and receiving information between the inside of the bias driving system (200) and between the bias driving system (200) and the outside.
[0020] The bias object information generation unit (250) includes a location recognition module (251), a road information fusion module (252), and an object fusion module (253).
[0021] The location recognition module (251) compares the recognition information received from the recognition sensor unit (210), the current location information of the vehicle received from the GPS receiver (230), and the precision map of the vehicle's surroundings received from the precision map transmission unit (220) to generate the current precision location information of the vehicle and location recognition reliability information. The road information fusion module (252) generates a precision map of the vehicle's surroundings using the current precision location information of the vehicle received from the location recognition module (251) and the precision map of the vehicle's surroundings received from the precision map transmission unit (220). The object fusion module (253) generates fused object information using the precision map of the vehicle's surroundings and the recognition information received from the road information fusion module (252). The fused object information is map information, but it is not simple map information; rather, it is map information that includes the location and speed of objects, namely surrounding vehicles other than the vehicle itself.
[0022] The control parameter generation unit (260) generates control parameters provided to the control unit (270) using fusion object information received from the bias object information generation unit (250), and detailed configuration and function will be described later.
[0023] The control unit (270) generates a control value of the vehicle corresponding to the control parameter received from the control parameter generation unit (260) and controls the deviation of the vehicle.
[0024] Figure 2 is an example of the configuration of a control parameter generation unit.
[0025] Referring to FIG. 2, the control parameter generation unit (260) includes a bias target object extraction module (261), an object virtual line extraction module (262), a bias object clustering module (263), a bias path generation module (264), a speed profile generation module (265), a driving path information generation module (266), and a control parameter generation module (267).
[0026] First, the details regarding the input and output of each component module will be briefly described, and the specific functions of each component module will be explained later.
[0027] The bias target object extraction module (261) extracts an object or a list of objects that cause bias in the self-vehicle among a plurality of objects included in the fused object information received from the bias object information generation unit (250). An object that causes bias in the self-vehicle refers to a vehicle that has approached within a pre-set bias interval (D).
[0028] The object virtual line extraction module (262) extracts a virtual line of the object causing the bias of the vehicle received from the bias target object extraction module (261), and extracts the center line of the vehicle using the extracted virtual line of the bias target object. Here, the virtual line is a virtual straight line (or curve) parallel to the vehicle lane based on the outermost point of the target object and the nearest point to the vehicle, and is used as a reference line during bias. When extracting the center line using the virtual line, the case where objects exist on both sides of the vehicle and the case where objects exist only on one side are distinguished, and in each case, the risk level, location recognition accuracy, driving tendency of the vehicle driver, road curvature, and road shape are reflected.
[0029] The bias object clustering module (263) generates bias cluster information by grouping the bias target objects according to the virtual line of the bias target objects received from the object virtual line extraction module (262) and a preset criterion. The bias cluster information includes the number of clusters and information of each cluster, and is used to prevent the generation of unnatural bias paths. Here, the criteria include a distance-based criterion, a zigzag driving-based criterion, and an estimated lateral speed-based criterion.
[0030] The bias path generation module (264) generates a bias path, which is a set of final bias values, using bias cluster information received from the bias object clustering module (263). Here, the bias path refers to a line connecting the center line of the virtual line of the left lane closest to the self lane and the virtual line of the right lane within the cluster, and the lane link.
[0031] The speed profile generation module (265) generates speed profile information including the speed information of the vehicle that matches the deviation path received from the deviation path generation module (264) and the traffic conditions of the road.
[0032] The driving path information generation module (266) generates driving path information including the bias path received from the bias path generation module (264) and the speed profile information received from the speed profile generation module (265).
[0033] The control parameter generation module (267) generates control parameters necessary for operating the vehicle according to the driving path information generated by the driving path information generation module (266).
[0034] Figure 3 illustrates the operation of the bias target object extraction module.
[0035] Referring to FIG. 3, it can be seen that the bias target object extraction module (261) includes only 2 objects (Object 2, Object 3) that are included in the bias section (D(m), m is a meter) among a total of 4 objects located around the vehicle as bias target objects. The bias section (D) is variably set according to the width of the lane and the characteristics of the lane.
[0036] As described above, the fused object information received by the bias target object extraction module (261) is a fusion of the location, speed, and map information of adjacent vehicles (objects 1 to 4) of the self-vehicle. The width of a typical lane is approximately 3.5m, for example, the distance to the lane of a vehicle with a total width of about 2m is approximately 50cm when considering a weak degree of bias, and since the lane is within 30cm and poses a threat to the opposite lane, the bias section (D) can be set by combining the area within 30cm outside the self-vehicle lane and the non-in-pass section within the self-vehicle lane.
[0037] Figure 4 illustrates the operation of the object virtual line extraction module.
[0038] Referring to FIG. 4, it can be seen that the object virtual line extraction module (262) extracts the virtual line of the bias target object (object 1, object 2) that causes the bias of the vehicle, received from the bias target object extraction module (261), not only on straight roads but also on curved roads. It has already been explained that the bias target object is a target vehicle or a list of target vehicles that the vehicle must bias.
[0039] In cases where the output of the various sensors constituting the perception sensor unit (210) is inaccurate and the outline of the vehicle shakes, an embodiment is also possible that uses a virtual line obtained by parallel shifting a certain width of the vehicle to the center of the rear bumper.
[0040] When objects (hereinafter referred to as "vehicles") exist on both sides of the road, it is desirable for the vehicle to travel in the exact center between the two vehicles; therefore, in this case, the center line can be calculated based on the outermost lines of the two vehicles. When a vehicle exists on only one side of the road, the center line can be calculated by assuming that an identical vehicle exists on the opposite side of the existing vehicle at the same distance.
[0041] After explaining how to determine the virtual line and the center line of the two virtual lines when vehicles are present on both roads, the method to determine the virtual line and the center line of the two virtual lines when vehicles are present on only one road will be explained.
[0042] Figure 5 is an example of a method for determining a virtual line and a center line when there are vehicles on both sides of the vehicle.
[0043] Referring to Fig. 5, it is clear that when the left vehicle (Object 1) and the right vehicle (Object 2) are driving along their respective center lines, the lane link and the center line will coincide. However, when both the left vehicle (Object 1) and the right vehicle (Object 2) are slightly deviated toward their respective lanes, it can be seen that, depending on the degree of deviation, the virtual center line that becomes the modified lane link—that is, the center point of the two vehicles (Object 1, Object 2)—is determined to be a certain distance to the left of the lane width of the two vehicles and the lane link. The more carefully the deviated vehicle (Object 2) is located in a section with a narrow lane width, the more the center line is deviated toward the opposite side, the left.
[0044] Valid deviation in the opposite direction can be ignored. That is, referring to Fig. 5, even if the width of the left lane relative to the vehicle's lane is wider than the width of the right lane, the vehicle does not move indefinitely toward that lane, and only the necessary degree of deviation is reflected.
[0045] The position of the virtual line of the vehicle on the left (Object 1) (P X0 , P Y ), position of the center line (Px, Py) and position of the virtual line of the vehicle on the right (Object 2) (P X1 The x-coordinate (Px) of the center line in , PY) can be expressed as Equation 1.
[0046]
[0047] Here, L1 represents the distance between the lane link of the vehicle's lane and the two lanes, L0 represents the distance between the vehicle's left lane and the center line of the vehicle on the left, and L2 represents the distance between the vehicle's right lane and the center line of the vehicle on the right.
[0048] D0 is the distance between the center line of the left lane and the imaginary line to the right of the left vehicle (Object 1) assuming the vehicle is traveling along the center line, and S0 is the distance traveled by the vehicle (Object 1) in the direction of its own lane, assuming it is traveling along the center line of the left lane. Alternatively, D0 can be represented as half the width of the left vehicle (Object 1).
[0049] D1 and S1 are for the right vehicle (Object 2) located in the right lane, and are applied in the same way as the concept applied to the left vehicle (Object 1).
[0050] Figure 6 illustrates a conventional method for determining a virtual line and a center line when a vehicle exists on only one side of the vehicle.
[0051] Referring to FIG. 6, when a vehicle (Object 2) exists only on the right side of the vehicle, the conventional method assumes that a vehicle (Object 1) is operating in the left lane of the vehicle with respect to the center line, and sets the virtual line on the left side of the vehicle to a point moved to the right by the width of the virtual Object 1 from the center line (or link lane) of the left lane. The center of the virtual line on the left set in this way and the virtual line on the right set in the manner shown in FIG. 5 is defined as the modified lane link of the vehicle as shown in Equation 2.
[0052]
[0053] However, the virtual line of the left lane, determined in the same way as mathematical formula 2, simply reflects that there are no vehicles in the left lane.
[0054] In the present invention, when there is no vehicle in one lane, the lane width of the lane without a vehicle can be changed according to the curvature of the lane without a vehicle, the shape of the lane, the risk level of the vehicle located on the other side (Object 2), the accuracy of the vehicle's position recognition, and the driving tendency of the vehicle's driver, thereby allowing the collision with the vehicle on the right (Object 2) that actually exists to be minimized by taking into account the vehicle on the left (Object 1) that actually exists.
[0055] FIG. 7 illustrates the method proposed in the present invention for determining a virtual line and a center line when a vehicle exists on only one side of the vehicle.
[0056] Referring to Fig. 7, when there are no vehicles in the left lane, only the possible scenarios involving vehicles on the right need to be considered; therefore, the width of the left lane is arbitrarily widened to minimize potential problems involving vehicles on the right. That is, since no vehicles are driving in the left lane, the center line (link crane) of the vehicle is moved further toward the left lane compared to the conventional method in order to maintain a sufficient distance between the vehicle and the vehicle on the right.
[0057] Since vehicles arbitrarily adjust the lane width of the existing lane, the position of the lane's center line also changes in proportion to the change in lane width. As the position of the virtual line in the left lane changes, the position of the center line also changes, and consequently, the vehicle's deflection value is also affected.
[0058] The new lane link of the vehicle, taking into account the left lane where the lane width changes, can be expressed as Equation 3.
[0059]
[0060] Referring to Equation 3, the parameters affecting the position of the virtual line are the lane width (L0'), the width of the virtual vehicle (D0), and the offset of the virtual vehicle (offset, L2-L0'). Therefore, it can be seen that if the concept of the virtual vehicle is excluded, the lane width can be changed.
[0061] Although the above description explained changes in lane width to change the position of the virtual line, the method for changing the virtual line is not limited to the above method. In particular, by introducing the concept of a virtual vehicle, the objective of changing the position of the virtual line can also be achieved through changes in the width and offset of the virtual vehicle.
[0062] Factors to consider when changing the width of a lane where vehicles are present include risk, location perception accuracy, the driving tendencies of the vehicle's owner, road curvature, and road shape.
[0063] The risk level refers to the accident risk based on the type and size of the vehicle, specifically the vehicle on the right. Since driving closer to a dangerous vehicle generally poses a greater threat to the driver, it is proposed to reflect this risk level when generating virtual lines for lanes where vehicles are absent.
[0064] Figure 8 illustrates a method for reflecting risk levels when setting virtual lines in the object virtual line extraction module.
[0065] Referring to FIG. 8, it can be seen that the object virtual line extraction module (262) can reflect the risk level when generating the virtual line of an object, and that the position of the virtual line of the left object is different depending on the type and size of the vehicle on the right. The driver of a vehicle feels a different sense of threat from surrounding vehicles depending on the type of vehicle operating in the vicinity. For example, the driver feels more threatened and drives cautiously when the surrounding vehicles are buses, construction vehicles, trucks, and motorcycles, rather than when the surrounding vehicles are small cars or passenger cars of the same size.
[0066] It is assumed that multiple lanes installed parallel to each other and capable of proceeding in the same direction have the same width, for example, that there are no vehicles in the left lane relative to the vehicle and there are vehicles in the right lane. In this case, when setting the driving path of the vehicle, the width of the left lane is arbitrarily expanded according to the type and size of the vehicle in the right lane, thereby shifting the driving path of the vehicle—that is, the vehicle's lane link—further toward the left lane, thereby preventing contact accidents between the vehicle and the vehicle on the right.
[0067] Referring to the upper part of Fig. 8, the width of the left lane is set to D1 when there is a less dangerous vehicle, such as a passenger car, in the right lane (Object 2-1). Referring to the lower part of Fig. 8, the width of the left lane is set to D2 when there is a dangerous vehicle, such as a cargo truck, in the right lane (Object 2-2). Here, D1 < D2.
[0068] Referring to Fig. 8, it can be seen that the width of the lane shows a weak bias when less dangerous vehicles are present, whereas it shows a strong bias when dangerous vehicles are present.
[0069] Setting a strong bias in the presence of such dangerous vehicles can ensure not only driving stability and ride comfort but also redundancy against autonomous driving control malfunctions.
[0070] Figure 9 illustrates a method for reflecting position recognition accuracy when setting virtual lines in the object virtual line extraction module.
[0071] Referring to FIG. 9, the object virtual line extraction module (262) can reflect the position recognition accuracy when generating the virtual line of an object, and proposes securing redundancy due to a misjudgment of the autonomous driving judgment logic by securing a wider gap with the vehicle in the adjacent lane when the position recognition is inaccurate. When vehicles are present in both lanes, the lane width in either direction is not changed, so no problem occurs in the operation of the vehicle. As shown in FIG. 9, when a vehicle is present only in the right lane, and the position recognition of the vehicle or the vehicle on the right is inaccurate, the width of the left lane is changed to be as wide as possible so that the final virtual line moves toward the left lane. That is, the degree of inaccuracy in the position recognition of the vehicle or the vehicle on the right is proportional to the width of the lane that is changed.
[0072] As shown in Fig. 8, D1 when there is a weak bias is narrower than D2 when there is a strong bias (D1 < D2).
[0073] Figure 10 illustrates a method for reflecting the driving tendencies of a vehicle driver when setting virtual lines in an object virtual line extraction module.
[0074] Referring to FIG. 10, the object virtual line extraction module (262) can reflect the driving tendencies of the driver of the vehicle on the road when generating the virtual line of the object, and supervises learning road conditions and object information, driver information and final bias values during manual driving. After learning is completed, the lane width change amount to derive the final bias value learned in a situation similar to the learned situation can be used as the basic offset, and additionally calculated offsets can be summed.
[0075] For example, for a driver driving with a deviation value (D1) of 10 cm when driving parallel to a passenger car in the right lane, an offset that changes the width of the left lane by 20 cm in the same situation can be applied by default.
[0076] If, in the same situation, the lane width needs to be changed by an additional 10 cm to reflect position recognition accuracy, the total lane width will be changed by 30 cm, resulting in a deviation value of 15 cm (D2).
[0077] Figure 11 illustrates a method for reflecting the driving tendencies of the vehicle driver when determining the lane width in the object virtual line extraction module.
[0078] Referring to FIG. 11, a method (1100) for reflecting the driving tendencies of a self-driving car driver includes a step (1110) of determining whether the self-driving car is currently operating in an autonomous driving mode, a step (1120~1140) of learning and storing driving tendencies for each driver when it is determined that it is not in an autonomous driving mode (1110, No), and a step (1150~1180) of determining biased driving based on the results of the steps (1120~1140) of learning and storing driving tendencies for each driver when it is determined that it is in an autonomous driving mode (1110, Yes) and map shape information.
[0079] The steps of learning and storing driving tendencies for each driver (1120~1140) include a step of storing the change in lane width of a lane adjacent to the lane currently being driven (1120), a step of performing lane width learning for each object (1130), and a step of storing the learning results for each driver (114).
[0080] The steps for determining biased driving (1150 to 1180) include: a step of determining whether existing learning results exist (1150); when it is determined that existing learning results exist (1150, Yes), a step of determining lane width by comparing the current situation with the learning results for each driver stored in the step of learning and storing driving tendencies for each driver (1120 to 1140); when it is determined that existing learning results do not exist (1150, No), a step of determining lane width using map shape information (1170); and a step of determining biased driving (1180) using the lane widths determined in the steps of determining two lane widths (1160, 1170).
[0081] Referring to FIG. 11, the bias driving system (200) according to the present invention can learn the driver's driving habits through the final bias value based on each object information and surrounding road shape when in manual mode rather than autonomous driving mode, and when in autonomous driving mode, if there is a previously learned model, the final bias value can be output as a learning result by inputting the surrounding road shape and object information and using it for bias.
[0082] In addition, when in autonomous driving mode and no existing learning model exists (initialization state,
[0083] (or in a state where the driver has switched, etc.) lane width determination based on map shape information can be performed without offset, and in autonomous driving mode when an existing learning model exists, lane width is determined based on the learning results so that an offset exists at the position of the virtual line, and bias driving can be performed by determining the final lane width and virtual line based on map shape information by reflecting the offset.
[0084] Figure 12 illustrates a method for reflecting the curvature of a road when setting a virtual line in the object virtual line extraction module.
[0085] Referring to FIG. 12, the object virtual line extraction module (262) can reflect the curvature of the road when generating the virtual line of the object, and when it is determined that there is curvature in the road, it can improve driving stability by adding a deviation in the direction of the center point rather than the exact center of the lane of the vehicle.
[0086] The deflection value is set to be proportional to the road curvature so that a larger deflection allows for a control margin of centrifugal force as the road curvature increases.
[0087] For example, when the curvature is positive, that is, when the vehicle turns left as shown in Fig. 12, a positive offset, that is, an offset to the left, is added to the virtual line, and when the curvature is negative, that is, when the vehicle turns right, a negative offset, that is, an offset to the right, is added to the virtual line.
[0088] Figure 13 illustrates a method for reflecting the shape of a road when setting a virtual line in the object virtual line extraction module.
[0089] Referring to FIG. 13, the object virtual line extraction module (262) can reflect the shape of the road when generating the virtual line of the object, and the shape of the road reflects additional bias for a vehicle (object 2) entering the road in a specific section such as an intersection or a merging section.
[0090] Since it is common for an entering vehicle (object 2) to drive along the shape of the road of the entrance or enter in a form similar to the shape of the road, the shape of the road of the entrance is reflected in the bias value. In order to reflect the road shape, the shape of the lane of the entering lane is reflected in the lane width of the opposite lane when the lane width changes, thereby easily generating a biased path that reflects this.
[0091] The red straight line on the upper left of Fig. 13 reflects the shape of the entrance road to the merging section, and the red curve on the lower left reflects the shape of the entrance road to the intersection.
[0092] Figure 14 is an example of a method for determining the final virtual line.
[0093] Referring to FIG. 14, it can be seen that the offset calculated by reflecting the risk level (1410), location recognition accuracy (1420), driver's driving tendency (1430), road curvature (1440), and road shape (1450), and the weights thereof are combined (1460), and the final virtual line can be determined by reflecting this.
[0094] The offset of the virtual line can be determined through weighted sum calculation by assigning weights proportional to the importance of each element, and the weight determination method can be determined experimentally through parameter tuning, or methods such as assigning a reward function and determining the parameter that maximizes the reward function through iterative feedback loop circulation can also be applied.
[0095] Although FIG. 14 illustrates the aggregation of all five calculation results, an embodiment reflecting at least one of them may also be possible.
[0096] FIG. 15 illustrates an example of clustering performed in a biased object clustering module.
[0097] Referring to FIG. 15, the bias object clustering module (263) groups multiple objects into one to minimize the generation of unnatural bias paths when the vehicle calculates the bias value of each object, and uses the distance between vehicles as one of the criteria for grouping.
[0098] In other words, as explained below, bias objects are grouped when the distance between them is smaller than the threshold distance.
[0099] - When the Nth bias object exists within a critical distance T(m) of the (N+1)th bias object, the two objects are grouped into one object.
[0100] - When the (N+1)th object exists within a critical distance T(m) of the (N+2)th bias object, objects N, (N+1), and (N+2) are grouped into a single object.
[0101] - The above-described method can be repeated for clustering up to a total of N_max, and N_max can be extended to the total number of biased objects.
[0102] - In the above situation, when determining the final bias path, the center line of the virtual line of the left lane closest to the self lane within the cluster and the virtual line of the right lane can be a component.
[0103] FIG. 16 illustrates another embodiment of clustering performed in the biased object clustering module.
[0104] Referring to FIG. 16, the bias object clustering module (263) groups multiple objects into one when the vehicle calculates the bias value of each object, and uses whether the other vehicle is driving in a zigzag pattern as another criterion for grouping.
[0105] In other words, as explained below, when there are multiple vehicles (objects) operating in a zigzag pattern, they are grouped into one.
[0106] - When the Nth bias object and the (N+1)th bias object exist in the left and right lanes, respectively, based on the ego lane, the two objects are grouped into a single cluster.
[0107] - When the (N+1)th object and the (N+2)th bias object exist in the left and right lanes, respectively, relative to the ego lane, objects N, (N+1), and (N+2) are grouped into a single cluster.
[0108] - The above method can be repeated for clustering up to a total of N_max, and N_max can be extended to the total number of biased objects.
[0109] - In the above situation, when determining the final bias path, the center line of the virtual line of the left lane closest to the self lane within the cluster and the virtual line of the right lane can be a component.
[0110] FIG. 17 illustrates another example of clustering performed in the biased object clustering module.
[0111] Referring to FIG. 17, the bias object clustering module (263) groups multiple objects into one when the vehicle calculates the bias value of each object, and uses the vehicle's expected lateral speed as another criterion for grouping.
[0112] In other words, as explained below, when the expected lateral speed of a vehicle is expected to be greater than the critical speed, multiple other vehicles (objects) corresponding to this are grouped together.
[0113] - When the Nth bias object and the (N+1)th bias object are mapped separately, the two objects are grouped into one cluster when the lateral expected speed of the vehicle is greater than the threshold speed.
[0114] - When the (N+1)th object corresponds separately to the (N+2)th bias object, the lateral expected speed of the vehicle is greater than the threshold speed, and the N, (N+1), and (N+2) objects are grouped into one cluster.
[0115] - The above method can be repeated for clustering up to a total of N_max, and N_max can be extended to the total number of biased objects.
[0116] - In the above situation, when determining the final bias path, the center line of the virtual line of the left lane closest to the self lane within the cluster and the virtual line of the right lane can be a component.
[0117] Bias clustering can be performed using one of the three methods mentioned above or by combining two or more of them in a complementary manner, and the output of the bias clustering module can be the total number of clusters and information for each cluster, and the components of the cluster information can be the number of biased objects within the cluster, information on biased objects within the cluster (velocity, location, map information, etc.), and virtual line information.
[0118] Figure 18 illustrates the bias path generated by the bias path generation module.
[0119] Referring to FIG. 18, it can be seen that the bias path generated by the bias path generation module (264) can be configured in the form of a curve that smoothly connects the center line of the virtual line of the left lane closest to the lane within the cluster and the virtual line of the right lane and the lane link.
[0120] The method of connecting the center line of a virtual line and the lane link is not limited to a specific method, and the following methods may serve as examples.
[0121] 1. Control the uniform points forming the centerline of the virtual line and the uniform points on the lane link
[0122] Bezier Curve using Points
[0123] 2. Control the uniform points forming the centerline of the virtual line and the uniform points on the lane link
[0124] B-spline curve using points
[0125] 3. Control the uniform points forming the centerline of the virtual line and the uniform points on the lane link
[0126] NURBS using points
[0127] 4. Control the uniform points forming the centerline of the virtual line and the uniform points on the lane link
[0128] Cubic spline using points
[0129] FIG. 19 is an example of a lane link and a road shape information-based deflection driving method according to the present invention.
[0130] Referring to FIG. 19, the lane link and road shape information-based bias driving method (1900) according to the present invention includes a step of initializing (1910), a step of extracting a bias target object (1920), a step of extracting a virtual line of a bias target object (1930), a step of clustering the bias targets (1940), a step of calculating a final bias offset (1950), a step of generating a bias path (1960), a step of generating a speed profile (1970), and a step of generating control parameters (1980).
[0131] Since the lane link and road shape information-based deflection driving method (1900) illustrated in FIG. 19 is performed using the lane link and road shape information-based deflection driving system (200) illustrated in FIG. 1 and FIG. 2, only the technical details performed at each step will be briefly explained.
[0132] In the initialization step (1910), the lane link and the road shape information-based deflection driving system (200) shown in FIG. 2 are initialized.
[0133] In the step (1920) of extracting the bias target object, the bias target object is extracted based on the bias interval.
[0134] In the step (1930) of extracting the virtual line of the object to be biased, the reference point and virtual line to be biased are extracted.
[0135] In the step of clustering bias targets (1940), bias objects that are close to the self-deviation are grouped according to a specific criterion.
[0136] In the step (1950) of calculating the final bias offset, the bias offset is finally determined for each grouped cluster.
[0137] In the step (1960) of generating a bias path, a path on the lane link that reflects the offset, i.e., a bias path that serves as a guideline for driving, is generated.
[0138] In the step of generating a speed profile (1970), a speed profile of the vehicle moving along the bias path is generated.
[0139] In the step of generating control parameters (1980), control parameters reflecting the bias path and velocity profile are generated.
[0140] Figure 20 illustrates an example of bias depending on the presence or absence of a vehicle in the opposite lane while driving in the outermost lane.
[0141] Referring to FIG. 20, when driving in the outermost lane, one may feel threatened when driving near bushes or guardrails and may drive in a biased direction. If there are no vehicles in the opposite lane, one can maintain sufficient distance from guardrails, bushes, etc., but if there are vehicles driving in the adjacent lane, one may generally maintain the center of the lane or drive in a biased direction toward the vehicles. This situation can be reflected in the bias value by assigning different risk levels to static objects such as bushes and guardrails and to vehicles in the adjacent lane.
[0142] For example, by making the change in lane width proportional to the risk of the opposing vehicle, as shown in the upper part of FIG. 20, the position of the final virtual line may be shifted toward the opposite side of the dangerous vehicle because the risk is greater than that of the bush when a dangerous vehicle is present. As shown in the lower part of FIG. 20, when no vehicle is present, the position of the final virtual line may be shifted toward the opposite side of the bush / guardrail because the risk of the bush / guardrail is higher than that of the empty lane.
[0143] FIG. 21 illustrates an example of the application of the present invention in a situation where dangerous vehicles and ordinary vehicles are mixed.
[0144] Referring to Fig. 21, when driving in a normal lane, when vehicles of different types are driving in both lanes, there is a tendency to drive with a bias, keeping a sufficient distance from the relatively more dangerous vehicle. This situation can be reflected in the bias value by assigning different risk levels to the dangerous vehicle and the relatively safe vehicle.
[0145] For example, by making the change in lane width proportional to the risk level of oncoming vehicles, the position of the final virtual line may be shifted toward the opposite side of the dangerous vehicle—that is, toward the less dangerous vehicle—since the presence of a dangerous vehicle carries a greater risk than that of a regular vehicle. Because the risk level of dangerous vehicles (trucks, construction vehicles, buses, etc.) is higher than that of regular vehicles, the final virtual line may be shifted toward the opposite side of the dangerous vehicle.
[0146] The present invention illustrated in FIGS. 11, 14, and 19 described above can be implemented as computer-readable code on a medium on which a program is recorded. A computer-readable medium includes all types of recording devices in which data that can be read by a computer system is stored. Examples of computer-readable media include HDD (Hard Disk Drive), SSD (Solid State Disk), SSD (Silicon Disk Drive), ROM, RAM, CD-ROM, magnetic tape, floppy disk, optical data storage device, etc.
[0147] Although the technical concept of the present invention has been described above together with the accompanying drawings, this is merely an illustrative explanation of preferred embodiments of the present invention and is not intended to limit the invention. Furthermore, it is evident that any person skilled in the art to which the present invention pertains can make various modifications and imitations within the scope of the technical concept of the present invention without departing from its scope. Explanation of the symbols
[0148] 210: Perception sensor unit 211: LiDAR 212: Camera 213: Radar 220: Precision Map Transmitter 230: GPS receiver 240: Communications Department 250: Bias object information generation section 260: Control Parameter Generation Unit 261: Bias Target Object Extraction Module 262: Object Virtual Line Extraction Module 263: Bias Object Clustering Module 264: Bias Path Generation Module 265: Speed Parameter Extraction Module 266: Driving Path Generation Module 267: Control Parameter Generation Module 270: Control unit
Claims
Claim 1 A control parameter generation unit that uses fused object information, which is map information including the location and speed of surrounding vehicles, to extract an object that causes bias of the vehicle, generates a virtual line of the extracted object by reflecting at least one of the risk level of surrounding vehicles, location recognition accuracy, driving tendency of the vehicle driver, road curvature, and road shape, and generates control parameters using the virtual line; The system comprises a control unit that generates a control value of a vehicle corresponding to a control parameter received from the control parameter generation unit and controls the deviation of the vehicle, wherein the control parameter generation unit includes: a deviation target object extraction module that extracts a deviation target object that causes the deviation of the vehicle among a plurality of objects included in the fused object information; and an object virtual line extraction module that extracts a virtual line of the deviation target object that causes the deviation of the vehicle and extracts the center line of the vehicle using the extracted virtual line of the deviation target object; and the object virtual line extraction module generates the virtual line by distinguishing between when objects exist on both sides of the vehicle and when objects exist only on one side, and is a lane link and a road shape information-based deviation driving system. Claim 2 In claim 1, the control parameter generation unit groups a plurality of objects from which the virtual line has been extracted using at least one of the following: a distance between vehicles, whether other vehicles are driving in a zigzag pattern, and the expected lateral speed of the self vehicle, and generates the control parameters using the grouped plurality of objects, a lane link and a road shape information-based bias driving system. Claim 3 In paragraph 2, the control parameter generation unit calculates the offset of each group, generates a deflection path by reflecting the calculated offset, generates a speed profile corresponding to the deflection path, and generates the control parameters by reflecting the deflection path and the speed profile, a lane link and a road shape information-based deflection driving system. Claim 4 In paragraph 3, the control parameter generation unit further comprises: a bias object clustering module that generates bias cluster information by grouping the bias target objects by applying at least one of the distance between vehicles, whether other vehicles are driving in a zigzag pattern, and the expected lateral speed of the self vehicle to the virtual line of the bias target objects; a bias path generation module that generates a bias path, which is a set of final bias values, using the bias cluster information; a speed profile generation module that generates speed profile information including the bias path and speed information of the self vehicle tailored to the traffic conditions of the road; a driving path information generation module that generates driving path information including the bias path and the speed profile information; and a control parameter generation module that generates the control parameters corresponding to the driving path information, thereby forming a lane link and a road shape information-based bias driving system. Claim 5 In paragraph 4, the object causing the deviation of the vehicle is a lane link and a road shape information-based deviation driving system, which means a vehicle that has approached within a preset deviation section. Claim 6 In paragraph 4, the above-mentioned bias cluster information includes a lane link and a road shape information-based bias driving system that includes clusters, i.e., the number of objects to be biased and information of each cluster. Claim 7 In paragraph 4, the above-mentioned deviation path refers to a lane link and a road shape information-based deviation driving system, wherein the lane link is a line connecting the center line of the virtual line of the left lane closest to the lane and the virtual line of the right lane. Claim 8 In paragraph 4, the above-mentioned virtual line is a lane link and a road shape information-based bias driving system, which is a virtual straight line or curve parallel to the lane of the target vehicle based on the outermost point of the target object and the nearest point of contact with the target vehicle. Claim 9 delete Claim 10 In paragraph 4, the above risk level is classified by the type and size of surrounding vehicles or by the magnitude of the sense of threat received from surrounding vehicles felt by the driver of the vehicle, a lane link and a road shape information-based bias driving system. Claim 11 In claim 8, the above position recognition accuracy includes at least one of the position recognition accuracy of the self-vehicle and the position recognition accuracy of surrounding vehicles, a lane link and a road shape information-based bias driving system. Claim 12 In paragraph 8, the driving tendency of the self-driving driver is a lane link and a road shape information-based bias driving system, which is the degree of habitual bias of the self-driving driver who drives separated from surrounding vehicles. Claim 13 In claim 8, a lane link and a road shape information-based deflection driving system in which the degree of deflection of the virtual line increases in proportion to the curvature of the road. Claim 14 In paragraph 8, the shape of the road is a lane link and a road shape information-based deflection driving system that distinguish whether the road is one of an intersection and a merging section, or whether the road is in the form of a straight or curved section. Claim 15 A lane link and a road shape information-based bias driving system, further comprising a bias object information generating unit that generates the fusion object information in claim 1. Claim 16 In claim 15, the bias object information generation unit comprises: a position recognition module that generates current precise location information of the vehicle by comparing recognition information received from the outside, current location information of the vehicle, and a precise map around the vehicle; a road information fusion module that generates a precise map around the vehicle using the current precise location information of the vehicle and the precise map; and an object fusion module that generates fused object information using the precise map around the vehicle and the recognition information, wherein the recognition information includes the driving road of the vehicle, surrounding information of the driving road of the vehicle, and information about vehicles around the driving road of the vehicle, a lane link and a road shape information-based bias driving system. Claim 17 A method for driving by bias based on a lane link and road shape information, comprising: a step of extracting a bias target object based on a bias section; a step of extracting a virtual line by reflecting a predetermined condition to a reference point to be biased and the extracted bias target object; a step of grouping objects corresponding to a predetermined standard among the bias objects close to the self-vehicle; a step of calculating a bias offset for each grouped object cluster; a step of generating a bias path, which is a path on a lane link reflecting the calculated offset; a step of generating a speed profile of the self-vehicle moving along the bias path; and a step of generating control parameters reflecting the bias path and the speed profile, wherein the step of extracting the virtual line distinguishes between when objects exist on both sides of the self-vehicle and when objects exist only on one side. Claim 18 In paragraph 17, the above-mentioned predetermined conditions include at least one of the risk level of surrounding vehicles, location recognition accuracy, driving tendency of the vehicle driver, road curvature, and road shape, and a lane link and a road shape information-based deflection driving method. Claim 19 In paragraph 17, the above-mentioned predetermined criteria include at least one of a lane link and a road shape information-based deflection driving method, wherein the lane link includes a distance-based method, a method based on whether other vehicles are driving in a zigzag pattern, and a method based on the expected lateral speed of the self vehicle. Claim 20 A computer-readable recording medium storing a program for executing the above-mentioned lane link and road shape information-based deflection driving method described in paragraph 17.
Citation Information
Patent Citations
Method creating drive course
KR1020110062439A
Apparatus and method for controlling drive of vehicle
KR1020190105154A
Method for road lane management based on vehicle-road infrastructure information fusion and apparatus for the same
KR1020200101516A
Travel control device and travel control method
WO2016031011A1
