Target screening method and device for intelligent driving and vehicle thereof

Abstract

The application discloses a target screening method and device for intelligent driving and a vehicle thereof. The target screening method for intelligent driving comprises the following steps: constructing a first current pipeline and a first prediction pipeline for each crossing target and each oblique crossing target; constructing a second current pipeline and a second prediction pipeline for each longitudinal moving target, each target moving from dynamic to static and each static target; calculating the predicted transverse position of each target within the TTC time; if the current position of the crossing target and the oblique crossing target is located in the first current pipeline and the predicted transverse position is located in the first prediction pipeline, the crossing target and the oblique crossing target are determined as effective targets; if the current position of the longitudinal moving target, the target moving from dynamic to static and the static target is located in the second current pipeline and the predicted transverse position is located in the second prediction pipeline, the longitudinal moving target, the target moving from dynamic to static and the static target are determined as effective targets. The application can screen out crossing or oblique crossing targets early, avoid risks, effectively screen out irrelevant targets and reduce the false screening rate.

Description

Technical Field

[0001] This invention relates to intelligent driving technology for automobiles. Background Technology

[0002] Figure 1 This diagram illustrates the control principle of longitudinal control functions in Advanced Driver Assistance Systems (ADAS), such as Adaptive Cruise Control (ACC), Automatic Emergency Braking (AEB), and Forward Collision Warning (FCW). The longitudinal control functions mainly include the following algorithm modules: visual perception and radar perception, target tracking and fusion, target selection, and vehicle control algorithms. The target selection module further includes the following functional modules: input target organization; road model construction; motion attribute judgment; selection pipeline construction; target selection for active safety functions (AEB, FCW, ESA, etc.) and comfort driving functions (ACC, TJA, NOA, etc.); rationality verification; confidence value output; and signal output.

[0003] The aforementioned "funnel" refers to a section of driving trajectory that starts from the front of the vehicle and extends outwards from both sides along the centerline of the vehicle's predicted driving trajectory, equal to a specified width. Targets falling within this driving trajectory area are considered to pose a danger to the vehicle's driving. Active safety systems (AEB, FCW, ESA, etc.) or comfort driving systems (ACC, TJA, NOA, etc.) in the ADAS / AD system will react to these targets (usually by braking). Targets outside the area are generally considered irrelevant and do not need to be considered by the system.

[0004] Figure 2 This diagram illustrates the principle behind active safety and comfort driving target selection. Please refer to it. Figure 2 Before the target selection module constructs the same pipeline for all targets, the motion attribute determination module receives all road participant targets output by the perception or fusion algorithm module. Based on the lateral movement speed, longitudinal speed, target type (pedestrian, vehicle, cyclist, etc.), and historical trajectory fitting information of each target, it assigns the motion attribute of each road participant to one of the following types:

[0005] 1. The default motion attribute MoveDir_Default = 0u;

[0006] 2. Move forward: MoveDir_Straight_Proceeding = 1u;

[0007] 3. Move to the left: MoveDir_Straight_Left = 2u;

[0008] 4. MoveDir_Straight_Oncoming = 3u;

[0009] 5. Move to the right: MoveDir_Straight_Right = 4u;

[0010] 6. MoveDir_Proceeding_Left = 5u;

[0011] 7. MoveDir_Oncoming_Left = 6u;

[0012] 8. Right rearward movement MoveDir_Oncoming_Right = 7u;

[0013] 9. Right forward motion MoveDir_Proceeding_Right = 8u;

[0014] 10. MoveDir_Stop = 9u (from motion to stillness);

[0015] 11. Never moved MoveDir_NBSM=10u.

[0016] Among them, targets with 1u and 3u attributes are longitudinally moving targets, targets with 2u and 4u attributes are horizontally crossing targets, targets with 5u, 6u, 7u, and 8u attributes are diagonally crossing targets, targets with 9u attributes are targets moving from motion to stillness, and targets with 10u attributes are stationary targets. The motion attribute judgment module will pass all target objects with assigned motion attributes to the downstream active safety target screening module and comfort driving target screening module. The aforementioned active safety target screening module includes, but is not limited to, the following target screening modules: AEB target screening module, FCW target screening module, ESA (Emergency Evasive Steering) target screening module, etc.; the aforementioned comfort driving target screening module includes, but is not limited to, the following target screening modules: ACC target screening module, TJA (Traffic Jam Assistant) target screening module, NOA (Navigate on Autopilot), etc.

[0017] In addition to receiving the target motion attributes determined by the upstream motion attribute judgment module, the active safety target selection module and the comfort driving target selection module also receive vehicle information obtained from the vehicle bus, including but not limited to: vehicle speed, yaw angle, and steering angle. Simultaneously, these modules also receive the target's physical attributes obtained from sensors or fusion algorithm modules, including the target's lateral and longitudinal positions, lateral and longitudinal velocities, acceleration, and the time to collision (TTC, i.e., the time remaining before the vehicle collides with the target object) and the time to brake (TTB) required to avoid the collision, calculated based on the vehicle and target's physical attributes. In addition, the relative position relationship module between the vehicle and the target calculates the following: the closest normal distance (XOLC) between the target and the vehicle's path at the current moment (lateral offset against lane center, meaning the lateral deviation of the target from the centerline of the vehicle's trajectory when transformed to the Frenet coordinate system); the predicted normal distance (Predict XOLC) of the target's expected collision position after the TTC time (Predict XOLC is the XOLC plus the predicted target position after the TTC time); the overlap ratio (Overlap) between the target's current position and the vehicle's trajectory (in ADAS / AD, this usually refers to the percentage of overlap between the vehicle and the target on the trajectory, considering the dimensions of both the vehicle and the target); and the predicted overlap ratio (Predict Overlap) between the target's future position and the vehicle. Using this information, the active safety target screening module and the comfort driving target screening module establish their own unique screening channels (funnels), and then perform target screening based on these channels.

[0018] The target screening channel currently used in the ADAS / AD industry is typically a channel that starts at the front of the vehicle, with a width equal to the lane width, and gradually narrows as it moves further away. This channel construction method has the following main drawbacks:

[0019] 1. The current target selection pipeline is chimney-shaped, with its width decreasing from near to far, typically at the width of a lane. It primarily targets vehicles traveling in the same or opposite directions, serving standard functions like adaptive cruise control (ACC) on highways. However, for complex urban scenarios involving road users such as pedestrians and cyclists (riding bicycles, electric bikes, etc.) who cross the road horizontally or diagonally, the existing pipeline construction method cannot select targets early enough. For example, for a cyclist crossing at 15 kph, the current target selection algorithm usually selects the target when it is still 13 meters laterally, leaving 3 seconds for downstream functions, resulting in insufficient time for timely braking to avoid risks.

[0020] 2. Most existing ADAS systems use visual and radar sensors for target perception. When processing targets on either side of the vehicle's trajectory that are very close longitudinally to the front of the vehicle, the detection accuracy usually decreases because the target is at the edge of the visual sensor and front radar's field of view, resulting in large errors in the lateral position and speed, leading to false positives. Furthermore, because the longitudinal distance is very close, these targets are easily identified as high-threat targets, causing the system to brake prematurely, resulting in a poor driving experience and even rear-end collisions. While a narrowing filter path, from near to far, uses a narrower selection criterion when processing distant targets to avoid false positives due to uncertainty caused by long processing times, it is highly prone to false positives when processing targets very close longitudinally to the front of the vehicle, especially VRUs (Vulnerable Road Users) near the vehicle's path.

[0021] 3. Furthermore, existing target screening methods, for a specific road participant, only consider the distance between the target and the vehicle when setting the screening channel width threshold. They apply the same screening channel standard to different types of road participants, neglecting the road participant's movement pattern, the vehicle's motion state, and the ADAS functions being served. Summary of the Invention

[0022] The technical problem to be solved by the present invention is to provide a target screening method, device and vehicle for intelligent driving, which can screen pedestrians, cyclists and other targets crossing or diagonally at an early stage to avoid risks, and can effectively screen out irrelevant targets in the vicinity of the vehicle in complex road scenarios, thereby reducing the false screening rate.

[0023] In a first aspect, a target selection method for intelligent driving according to embodiments of the present invention includes:

[0024] Construct a first current pipeline and a first predicted pipeline for each traversing target and each diagonally traversing target;

[0025] Construct a second current pipeline and a second prediction pipeline for each longitudinally moving target, each target moving from motion to stillness, and each stationary target;

[0026] Calculate the predicted lateral position of each target within the TTC time. If the current position of a traversing target or a diagonal target is within the first current channel, and its predicted lateral position is also within the first prediction channel, then the traversing target or diagonal target is determined to be a valid target. If the current position of a longitudinally moving target, a target moving from motion to stillness, or a stationary target is within the second current channel, and its predicted lateral position is also within the second prediction channel, then the longitudinally moving target, target moving from motion to stillness, or stationary target is determined to be a valid target. Wherein:

[0027] The constructed first current pipeline has the following shape: from the front of the vehicle forward, it consists of a widening section, a constant width section, a narrowing section, and a final section. The width of the widening section gradually increases from one end at the front of the vehicle to the other. The width w0 of the widening section at the front of the vehicle is 0.5W to 1.5W, where W is the width of the vehicle. The width w1 of the other end of the widening section is: w1 = V_lat * TTC + offset1, where V_lat is the lateral movement speed of the target, TTC is the collision time of the target, and offset1 is a preset compensation value, offset1 ≥ 0. The width of the constant width section is the same as the width of the other end of the widening section, and the width of the constant width section remains unchanged. The width of the narrowing section gradually decreases from one end connecting to the constant width section to the other end. The width w3 of the other end of the narrowing section is: w3 = 0.6W to... W; The width of the final segment is the same as the width of the other end of the narrowing segment, and the width of the final segment remains unchanged; The distance d1 between the other end of the widening segment and the front of the vehicle is set as: d1 = x - offset2, and the distance d2 between one end of the narrowing segment and the front of the vehicle is set as: d2 = x + offset2. When selecting targets for active safety functions, x is the braking distance or alarm distance. When selecting targets for comfort driving functions, x is the braking distance, and offset2 is a preset compensation value, offset2 > 0; The distance d3 between the other end of the narrowing segment and the front of the vehicle is set as: d3 = x + offset2 + k1 * Vr, where Vr is the longitudinal relative speed between the target and the vehicle, k1 is a preset empirical value, k1 = 0.2 seconds to 1 second, and the distance between the other end of the final segment and the front of the vehicle is set to infinity;

[0028] The constructed second current pipe has the following shape: from the front of the vehicle forward, it consists of a gradually narrowing section and a constant-width end; the width of the gradually narrowing section gradually narrows from one end at the front of the vehicle to the other end, with the width w4 at one end at the front of the vehicle being 0.7W to 1.1W, and the width w5 at the other end of the gradually narrowing section being 0.6W to W; the width of the constant-width end is the same as the width of the other end of the gradually narrowing section, and the width of the constant-width section remains unchanged; the distance from the other end of the gradually narrowing section to the front of the vehicle is set as d3', and the calculation formula for d3' is the same as that for d3; the distance from the other end of the constant-width section to the front of the vehicle is set to infinity;

[0029] The constructed first and second prediction pipes both have the following shapes: from the front of the vehicle forward, they consist of a tapering section and a constant-width section; the width of the tapering section gradually narrows from one end at the front of the vehicle to the other end, with the width w6 at one end of the tapering section being 0.8W to 1.2W, and the width at the other end of the tapering section being equal to w7, which is 0.3W to 1.1W; the width of the constant-width section is the same as the width at the other end of the tapering section, and the width of the constant-width section remains unchanged; the distance from the other end of the tapering section to the front of the vehicle is set as d3", and the calculation formula for d3 is the same as that for d3, while the distance from the other end of the constant-width section to the front of the vehicle is set to infinity.

[0030] Secondly, according to embodiments of the present invention, a target selection method for intelligent driving includes:

[0031] Construct a first current pipeline and a first predicted pipeline for each traversing target and each diagonally traversing target;

[0032] Calculate the predicted lateral position of each traversing target and each diagonal crossing target within the TTC time; if the current position of a traversing target or a diagonal crossing target is within the first current conduit, and the predicted lateral position is within the first predicted conduit, then the traversing target or diagonal crossing target is determined to be a valid target; where:

[0033] The constructed first current pipeline has the following shape: from the front of the vehicle forward, it consists of a widening section, a constant width section, a narrowing section, and a final section. The width of the widening section gradually increases from one end at the front of the vehicle to the other. The width w0 of the widening section at the front of the vehicle is 0.5W to 1.5W, where W is the width of the vehicle. The width w1 of the other end of the widening section is: w1 = V_lat * TTC + offset1, where V_lat is the lateral movement speed of the target, TTC is the collision time of the target, and offset1 is a preset compensation value, offset1 ≥ 0. The width of the constant width section is the same as the width of the other end of the widening section, and the width of the constant width section remains unchanged. The width of the narrowing section gradually decreases from one end connecting to the constant width section to the other end. The width w3 of the other end of the narrowing section is: w3 = 0.6W to... W; The width of the final segment is the same as the width of the other end of the narrowing segment, and the width of the final segment remains unchanged; The distance d1 between the other end of the widening segment and the front of the vehicle is set as: d1 = x - offset2, and the distance d2 between one end of the narrowing segment and the front of the vehicle is set as: d2 = x + offset2. When selecting targets for active safety functions, x is the braking distance or alarm distance. When selecting targets for comfort driving functions, x is the braking distance, and offset2 is a preset compensation value, offset2 > 0; The distance d3 between the other end of the narrowing segment and the front of the vehicle is set as: d3 = x + offset2 + k1 * Vr, where Vr is the longitudinal relative speed between the target and the vehicle, k1 is a preset empirical value, k1 = 0.2 seconds to 1 second, and the distance between the other end of the final segment and the front of the vehicle is set to infinity;

[0034] The constructed first predictive pipeline has the following shape: from the front of the vehicle forward, it consists of a tapering section and a constant-width section; the width of the tapering section gradually narrows from one end at the front of the vehicle to the other end, with the width w6 at the front end of the tapering section being 0.8W to 1.2W, and the width at the other end of the tapering section being equal to w7, which is 0.3W to 1.1W; the width of the constant-width section is the same as the width at the other end of the tapering section, and the width of the constant-width section remains unchanged; the distance from the other end of the tapering section to the front of the vehicle is set as d3", and the calculation formula for d3 is the same as that for d3, while the distance from the other end of the constant-width section to the front of the vehicle is set to infinity.

[0035] Thirdly, according to an embodiment of the present invention, a target screening device for intelligent driving includes at least one processor and a memory communicatively connected to the at least one processor; the memory stores instructions executable by the at least one processor; the at least one processor is used to execute the instructions to implement the aforementioned target screening method.

[0036] Fourthly, the vehicle according to an embodiment of the present invention includes the target screening device described above.

[0037] This invention has at least the following advantages and features:

[0038] 1. The target screening method of this invention improves the reliability and accuracy of screening targets that cross or diagonally by constructing current and predicted pipes of specific shapes for targets that cross or diagonally.

[0039] 2. Based on multiple dimensions such as the target's position and speed information, the target's motion mode, the vehicle's motion state, and the intelligent driving functions it serves, the present invention customizes a target filtering pipeline for each target in real time, thereby improving the overall accuracy of the filtering results. Attached Figure Description

[0040] Figure 1 A schematic diagram illustrating the longitudinal control function of an Advanced Driver Assistance System (ADAS) is shown.

[0041] Figure 2 A schematic diagram illustrating the principle of active safety and comfort driving target selection is shown.

[0042] Figure 3 A flowchart illustrating a target selection method for intelligent driving according to an embodiment of the present invention is shown.

[0043] Figures 4 to 7 The diagrams schematically illustrate the shapes of the first current pipeline, the first predicted pipeline, the second current pipeline, and the second predicted pipeline constructed according to embodiments of the present invention.

[0044] Figure 8 A schematic diagram of a scenario according to a specific embodiment of the present invention is shown. Detailed Implementation

[0045] Figure 3 A flowchart illustrating a target selection method for intelligent driving according to an embodiment of the present invention is shown. The intelligent driving described in this embodiment includes, but is not limited to, Advanced Driver Assistance Systems (ADAS) and autonomous driving. Please refer to... Figure 3 According to an embodiment of the present invention, a target selection method for intelligent driving includes the following steps.

[0046] a. Construct the first current pipeline and the first predicted pipeline for each traversing target and each diagonally crossing target.

[0047] Please refer to Figure 4The first current pipe 11 constructed has the following shape: from the front of the vehicle forward, it consists of a widening section 11a, a constant width section 11b, a narrowing section 11c, and a final section 11d. The width of the widening section 11a gradually increases from one end at the front of the vehicle to the other. The width w0 of the widening section 11a at the front of the vehicle is 0.5W to 1.5W, where W is the width of the vehicle itself. The width w1 of the other end of the widening section 11a is: w1 = V_lat * TTC + offset1, where V_lat is the lateral movement speed of the crossing target. For pedestrians, cyclists, and vehicles crossing the road, V_lat is relatively large, so the calculated w1 is definitely greater than w0. TTC is the collision time of the crossing target, and offset1 is a preset compensation value, offset1 ≥ 0. The width of the constant width section 11b is the same as the width of the other end of the widening section 11a, and the width of the constant width section remains unchanged. The width of the narrowing section 11c gradually decreases from one end connecting to the constant width section to the other. The width w3 of the other end of the narrowing section 11c is: w3 = 0.6W to W. The width of segment 11d is the same as the width of the other end of the narrowing segment 11c, and the width of the final segment remains unchanged; the distance d1 between the other end of the widening segment 11a and the front of the vehicle is set as: d1 = x - offset2, and the distance d2 between the one end of the narrowing segment 11c and the front of the vehicle is set as: d2 = x + offset2. When selecting targets for active safety functions, x is the braking distance or alarm distance. When selecting targets for comfort driving functions, x is the braking distance, and offset2 is a preset compensation value, offset2 > 0; the distance d3 between the other end of the narrowing segment and the front of the vehicle is set as: d3 = x + offset2 + k1 * Vr, where Vr is the longitudinal relative speed between the target (here, the target crossing horizontally or diagonally) and the vehicle, that is, the difference between the longitudinal movement speed of the target and the speed of the vehicle, k1 is a preset empirical value, k1 = 0.2 to 1 second, and the distance between the other end of the final segment and the front of the vehicle is set to infinity.

[0048] In some specific implementations, w0 is set to 0.8W to 1.2W. Active safety functions include one of the following: Automatic Emergency Braking (AEB), Forward Collision Warning (FCW), and Emergency Steering Assist (ESA); comfort driving functions include one of the following: Adaptive Cruise Control (ACC), Traffic Jam Assist (TJA), and Navigation-Assisted Driving (NOA).

[0049] Braking distance and warning distance are determined by the active safety functions or comfort driving functions served by the target screening system, and are parameters that can be obtained externally. The formula for calculating braking distance is: v target Let v be the target's velocity relative to the ground. ego'x' represents the vehicle speed, and 'a' represents the deceleration determined by the active safety function or comfort driving function. When the active safety function is Automatic Emergency Braking (AEB), 'x' represents the braking distance. When the active safety function is Forward Collision Warning (FCW) or Emergency Steering Assist (ESA), 'x' represents the warning distance, which is a preset value.

[0050] In this embodiment, offset2 is equal to k2*Vr or a preset constant, and k2 is equal to 0.1 seconds to 0.5 seconds.

[0051] like Figure 5 As shown. The constructed first prediction pipe 12 has the following shape: from the front of the vehicle forward, it includes a tapering section 12a and a constant-width section 12b. The width of the tapering section 12a gradually narrows from one end at the front of the vehicle to the other end. The width w6 of the tapering section 12a at the front of the vehicle is 0.8W to 1.2W, and the width of the other end of the tapering section 12a is equal to w7, which is 0.3W to 1.1W. The width of the constant-width section 12b is the same as the width of the other end of the tapering section, and the width of the constant-width section 12b remains unchanged. The distance from the other end of the tapering section 12a to the front of the vehicle is set as d3". The calculation formula of d3" is the same as that of d3, that is, d3" = x + offset2 + k1 * Vr, where x is the braking distance. The distance to the alarm is set, offset2 is a preset compensation value, Vr is the longitudinal relative speed between the target (here, the target crossing horizontally or diagonally) and the vehicle, k1 is a preset empirical value, k1 = 0.2 seconds to 1 second, and the distance from the other end of the equal-width section to the front of the vehicle is set to infinity. The width of the tapered section 12a at the front of the vehicle can be equal to or unequal to the width of the widening section 11a at the front of the vehicle. It should be noted that although the calculation formula for d3” is the same as that for d3, the values ​​of offset2 and k1 can be set differently according to the actual situation.

[0052] b. Construct a second current pipeline and a second prediction pipeline for each longitudinally moving target, each target moving from motion to stillness, and each stationary target.

[0053] like Figure 6As shown. The constructed second current pipe 21 has the following shape: from the front of the vehicle forward, it consists of a gradually narrowing section 21a and a uniform width end 21b. The width of the narrowing section 21a gradually narrows from one end at the front of the vehicle to the other. The width w4 of the narrowing section 21a at the front of the vehicle is 0.7W to 1.1W, and the width w5 of the other end of the narrowing section 21a is 0.6W to W. The width of the same-width end 21b is the same as the width of the other end of the narrowing section 21a, and the width of the same-width section remains unchanged. The distance between the other end of the narrowing section 21a and the front of the vehicle is equal to d3'. The calculation formula of d3' is the same as that of d3, that is, d3'=x+offset2+k1*Vr, where x is the braking distance or warning distance, offset2 is the preset compensation value, Vr is the longitudinal relative speed between the target (here, a longitudinally moving target, a target moving from motion to stillness, or a stationary target) and the vehicle, and k1 is a preset empirical value, k1=0.2 seconds to 1 second. The distance between the other end of the same-width section 21b and the front of the vehicle is set to infinity. Optionally, the width of the narrowing segment 21a at one end of the vehicle front is narrower than the width of the widening segment 11a at one end of the vehicle front. It should be noted that although the calculation formula for d3' is the same as that for d3, the values ​​of offset2 and k1 can be set differently depending on the actual situation.

[0054] like Figure 7 As shown. The constructed second predictive pipe 22 has the following shape: from the front of the vehicle, it consists of a tapering section 22a and a constant-width section 22b. The width of the tapering section 22a gradually narrows from one end at the front of the vehicle to the other end. The width w6 of the tapering section 22a at the front of the vehicle is 0.8W to 1.2W, and the width of the other end of the tapering section 22a is equal to w7, which is 0.3W to 1.1W. The width of the constant-width section 22b is the same as the width of the other end of the tapering section 22a, and the width of the constant-width section remains unchanged. The distance from the other end of the tapering section 22a to the front of the vehicle is set as d3”. The calculation formula for d3” is the same as that for d3, that is, d3” = x + offse t2+k1*Vr, where x is the braking distance or alarm distance, offset2 is the preset compensation value, Vr is the longitudinal relative speed between the target (here, a longitudinally moving target, a target moving from motion to stillness, or a stationary target) and the vehicle, k1 is a preset empirical value, k1 = 0.2 seconds to 1 second, and the distance from the other end of the equal-width section 22b to the front of the vehicle is set to infinity. It should be noted that although the calculation formula for d3” is the same as d3, the values ​​of offset2 and k1 can be set differently according to the actual situation. Furthermore, the values ​​of offset2 and k1 in the calculation formula for d3” of the second prediction pipe can also be set differently from those in the calculation formula for d3” of the first prediction pipe, depending on the actual situation.

[0055] c. Calculate the predicted lateral position of each target within the TTC time. If the current position of a traversing target or a diagonal target is within the first current channel 11 and the predicted lateral position is within the first prediction channel 12, then the traversing target or diagonal target is determined to be a valid target. If the current position of a longitudinally moving target, a target moving from motion to stillness, or a stationary target is within the second current channel 21 and the predicted lateral position is within the second prediction channel 22, then the longitudinally moving target, a target moving from motion to stillness, or a stationary target is determined to be a valid target.

[0056] The predicted horizontal position is calculated as V_lat * TTC + current position, where TTC is a variable.

[0057] The timing of steps a and b is not limited to the order of the textual descriptions mentioned above. For example, step b may precede step a.

[0058] The target selection method in this embodiment determines whether the current lateral position of the target falls in the current pipe ahead of the vehicle's driving trajectory, and also determines whether the target will continue to fall in another predicted pipe of a different shape in the future. When both conditions are met, the target can be selected as the driving target of the intelligent driving function.

[0059] This embodiment's method constructs a customized filtering pipeline for each road participant within the field of view. Based on multiple dimensions such as the obtained target's position and velocity information, target movement mode, vehicle's motion state, and the intelligent driving functions served, it customizes the target filtering pipeline for each target in real time, thereby improving the overall accuracy of the filtering results. The target's position information includes its lateral and longitudinal position in the Frenet coordinate system. The target's velocity information includes its lateral and longitudinal velocities, acceleration, and the collision time (TTC) and braking time (TTB) required to avoid the collision, calculated based on the vehicle's and target's physical properties. The motion attribute assignments to the target represent its movement mode: targets moving to the left and right are classified as traversing targets; targets moving to the left front, left rear, right front, and right front are classified as diagonally traversing targets; targets moving forward and towards each other are classified as longitudinally moving targets. Additionally, there are targets moving from motion to stillness and stationary (never moved) targets. The vehicle's motion state includes information such as its speed. The intelligent driving function being served refers to which intelligent driving function of this vehicle is being selected for the current target screening, such as AEB, FCW, ESA, ACC, TJA, NOA, etc. Different intelligent driving functions have different physical parameters (such as vehicle speed, vehicle acceleration, etc.).

[0060] The following combination Figure 8 First Figure 8Taking Target2 (crossing pedestrians) as an example, this illustrates the construction process of the AEB function filtering pipeline.

[0061] Target2 is moving laterally at its current location in the diagram with a lateral movement speed of V_lat. Since it is a traversing target, a first current pipe is constructed for it. At a distance of 0 from the front bumper of the vehicle (i.e., the front of the car), and... Figure 8 At point dx0 (which is equal to 0), the first current pipe ( Figure 8 The width w0 (shown as a solid line in the middle) is the width of this vehicle, extending forward to the other end of the gradually widening section (and...). Figure 8 During the process of dx1 remaining constant, the width gradually increases to its maximum value w1. The sizes of d1 and d2 in the first current pipe are related to the function of the target service being filtered. For the AEB function, the allowable deceleration magnitude 'a' is known prior information, obtained through the formula... The braking distance x is calculated. If target selection is performed for Adaptive Cruise Control (ACC), the x obtained using the same calculation method will be larger than the x obtained for AEB due to the smaller deceleration of ACC. Similarly, we can use prior experience to determine the magnitude of x for target selection in other driver assistance functions such as Forward Collision Warning (FCW). To compensate for uncertainties in the process and increase the robustness of the selection algorithm, based on the calculated x value, we increase and decrease the compensation distance by a factor of offset2 to obtain the magnitudes d1 and d2 respectively. The value of offset2 is set as the relative speed Vr between Target2 and the vehicle multiplied by 0.2 seconds, or simplified to a constant value such as approximately 5 meters.

[0062] The width w1 at the other end of the gradually widening section is calculated using the formula V_lat*TTC+offset1. offset1 is set to a constant value, which in this embodiment is half the vehicle width. This offset1 is added to control for missed potential threats due to measurement inaccuracies caused by sensor errors. Using this w1 value as a filtering threshold for whether a target's lateral position falls within the vehicle's trajectory ensures that targets not yet within this range (e.g., [missing information]) are not currently within this range. Figure 8 Target1 (assuming its V_lat and offset1 are the same as Target2) has only a very small probability of reaching the collision zone with its own vehicle within the future TTC time. Conversely, a target (such as...) Figure 8If Target2) is currently within the gradually widening section of the first current pipe, it has a high probability of being hit by the vehicle. After obtaining the w1 value of the first current pipe, the target screening method of this embodiment ensures that, at this relative speed, targets whose longitudinal distance falls within the constant width section of the first current pipe are all screened using the width of w1 as the screening threshold for determining whether the target falls within the vehicle's trajectory.

[0063] The reason why the widening section of the first-stage pipeline is designed in a gradually expanding trumpet shape is primarily to compensate for the unpredictability of the movement of nearby targets. Figure 8 Taking Target1 as an example, it moves across the vehicle at the same lateral speed. However, as the target gets closer to the front of the vehicle, the Time To Catch (TTC) decreases, meaning the vehicle will overtake the target in a shorter time. Correspondingly, the probability of Target1 suddenly entering the vehicle's path within the corresponding TTC decreases. Therefore, the width of the gradually widening section gradually increases from the vehicle width to a w1 shape, reducing false positives for surrounding targets. Since Target1's current position is outside the first current pipeline, it will not be selected as a valid AEB target.

[0064] The other end of the first narrowing section of the current pipe (and) Figure 8 The distance d3 from the front of the vehicle (dx3 is constant) is set as: d3 = d2 + 0.5 * Vr, where 0.5 seconds is an empirical value and can fluctuate during actual deployment. The width of the narrowing section decreases sharply from w2 (w2 = w1) to w3, where w3 is 0.8 times the width W of the vehicle. A very strict width is used as the threshold because, firstly, even if targets in this area are selected in time, it would still be outside the working range of the AEB function; secondly, for more distant targets, due to the larger TTC, the targets exhibit greater randomness in movement within a larger time window, and a conservative screening strategy is beneficial for controlling false screenings. Continuing forward from the other end of the narrowing section to infinity, the lateral width of the pipe remains constant at w3.

[0065] To ensure that Target2 does not exceed or leave its vehicle's trajectory after the future TTC time, the predicted lateral position of Target2 during the TTC time is also calculated synchronously in real time. The size of this predicted lateral position is compared with the width threshold of the first prediction channel to see if it falls within the first prediction channel. (First prediction channel...) Figure 8The area enclosed by the black dashed line (d3") starts from the front of the vehicle and extends outwards. Its width begins at the front and is slightly wider than the vehicle's width (e.g., 1.1 times the vehicle's width), gradually decreasing to 0.5 times the vehicle's width as the distance increases to d3". This 0.5 times is also an empirical value. The width of the first prediction pipe is usually determined using linear interpolation or slight adjustments, resulting in a straight line or a slightly modified straight line as shown in the diagram.

[0066] Considering both the current and future positions of the target, since Target2's current position falls within the first current pipeline and its predicted position falls within the first prediction pipeline, Target2 is selected as a valid AEB target.

[0067] Next, taking Target3 as an example, we will further explain the principle of customizing the screening pipeline for each participant in the road in this embodiment. After the motion state attribute judgment module has made its judgment, Target3's motion attribute is forward movement. For a target with such a motion attribute, the construction of the current position judgment pipeline will be different from that of Target2. The second current pipeline constructed for it is as follows: Figure 4 The dotted lines in the diagram indicate this. The second current pipeline starts at the front of the vehicle, where its width is 0.9 times the width of the vehicle body. At a distance d3' from the front, the width converges to half the width of the vehicle body. Linear interpolation can be used from the front to the other end of the narrowing section, resulting in a straight line as shown in the diagram. Alternatively, slight width adjustments can be made based on the scene. Clearly, Target3 is not currently within the second current pipeline. Since Target3 is moving longitudinally (with a longitudinal movement speed of V_long), its predicted position also falls outside the second prediction pipeline. Therefore, Target3 will not be selected as an AEB target of interest, and the AEB function will not perform any action on Target3.

[0068] This invention also provides a target screening device for intelligent driving, including at least one processor and a memory communicatively connected to the at least one processor; the memory stores instructions executable by the at least one processor; the at least one processor is used to execute the instructions to implement the aforementioned target screening method.

[0069] Obviously, those skilled in the art can make various modifications and variations to this invention without departing from its spirit and scope. Therefore, if these modifications and variations fall within the scope of the claims of this invention and their equivalents, this invention also intends to include these modifications and variations.

Claims

1. A target selection method for intelligent driving, characterized in that, include: Construct a first current pipeline and a first predicted pipeline for each traversing target and each diagonally traversing target; Construct a second current pipeline and a second prediction pipeline for each longitudinally moving target, each target moving from motion to stillness, and each stationary target; Calculate the predicted lateral position of each target within the TTC time. If the current position of a traversing target or a diagonal target is within the first current channel, and its predicted lateral position is also within the first prediction channel, then the traversing target or diagonal target is determined to be a valid target. If the current position of a longitudinally moving target, a target moving from motion to stillness, or a stationary target is within the second current channel, and its predicted lateral position is also within the second prediction channel, then the longitudinally moving target, target moving from motion to stillness, or stationary target is determined to be a valid target. Wherein: The constructed first current pipe has the following shape: from the front of the vehicle forward, it consists of a gradually widening section, a constant width section, a gradually narrowing section, and a final section; the width of the gradually widening section gradually increases from one end at the front of the vehicle to the other end, and the width w0 of the gradually widening section at the front of the vehicle is 0.5W to 1.5W. W represents the width of the vehicle. The width w1 at the other end of the widening section is: w1 = V_lat * TTC + offset1, where V_lat is the target's lateral movement speed, TTC is the target's collision time, and offset1 is a preset compensation value, offset1 ≥ 0. The width of the constant width section is the same as the width at the other end of the widening section, and the width of the constant width section remains unchanged. The width of the narrowing section gradually narrows from one end connecting to the constant width section to the other end, and the width w3 at the other end of the narrowing section is: w3 = 0.6W ~ W. The width of the final section is the same as the width at the other end of the narrowing section, and the width of the final section remains unchanged. The distance d1 between the other end of the widening section and the front of the vehicle is set as: d1 = x - offset2, and the distance d2 between the one end of the narrowing section and the front of the vehicle is set as: d2 = x + offset2. When selecting targets for active safety functions, x is the braking distance or warning distance. When selecting targets for comfort driving functions, x is the braking distance, offset2 is a preset compensation value, offset2>0. The distance d3 between the other end of the narrowing section and the front of the vehicle is set as: d3=x+ offset2+k1*Vr, where Vr is the longitudinal relative speed between the target and the vehicle, k1 is a preset empirical value, k1=0.2 seconds to 1 second, and the distance between the other end of the final section and the front of the vehicle is set to infinity. The constructed second current pipe has the following shape: from the front of the vehicle forward, it consists of a gradually narrowing section and a constant-width end; the width of the gradually narrowing section gradually narrows from one end at the front of the vehicle to the other end, with the width w4 at one end at the front of the vehicle being 0.7W to 1.1W, and the width w5 at the other end of the gradually narrowing section being 0.6W to W; the width of the constant-width end is the same as the width of the other end of the gradually narrowing section, and the width of the constant-width section remains unchanged; the distance from the other end of the gradually narrowing section to the front of the vehicle is set as d3', and the calculation formula for d3' is the same as that for d3; the distance from the other end of the constant-width section to the front of the vehicle is set to infinity; The constructed first and second prediction pipes both have the following shapes: from the front of the vehicle, they consist of a tapering section and a constant-width section. The width of the tapering section gradually narrows from one end at the front of the vehicle to the other end. The width w6 of the tapering section at the front of the vehicle is 0.8W to 1.2W, and the width w7 of the other end of the tapering section is equal to w7, which is 0.3W to 1.1W. The width of the constant-width section is the same as the width of the other end of the tapering section, and the width of the constant-width section remains unchanged. The distance between the other end of the tapering section and the front of the vehicle is set as d3'', and the calculation formula for d3'' is the same as that for d3. The distance between the other end of the constant-width section and the front of the vehicle is set to infinity.

2. The target screening method according to claim 1, characterized in that, The active safety function is one of the following: Automatic Emergency Braking (AEB), Forward Collision Warning (FCW), and Emergency Steering Assist (ESA); The comfort driving function is one of the following: Adaptive Cruise Control (ACC), Traffic Jam Assist (TJA), and Navigation Assist (NOA).

3. The target screening method according to claim 2, characterized in that, When the active safety function is Automatic Emergency Braking (AEB), x represents the braking distance; when the active safety function is Forward Collision Warning (FCW) or Emergency Steering Assist (ESA), x represents the warning distance.

4. The target screening method according to claim 1 or 3, characterized in that, The formula for calculating the braking distance x is: v target Let v be the target's velocity relative to the ground. ego denoted as the vehicle speed, and 'a' as the deceleration determined by the active safety function or the comfort driving function.

5. The target screening method according to claim 1, characterized in that, offset2 equals k2*Vr, where k2 equals 0.1 seconds to 0.5 seconds.

6. The target screening method according to claim 1, characterized in that, The width of the narrowing section at one end of the vehicle front is narrower than the width of the widening section at one end of the vehicle front.

7. A target screening device for intelligent driving, comprising at least one processor and a memory communicatively connected to the at least one processor; the memory storing instructions executable by the at least one processor; characterized in that, The at least one processor is used to execute the instructions to implement the method as described in any one of claims 1 to 6.

8. A vehicle, characterized in that, Includes the target screening device as described in claim 7.

9. A target selection method for intelligent driving, characterized in that, include: Construct a first current pipeline and a first predicted pipeline for each traversing target and each diagonally traversing target; Calculate the predicted lateral position of each traversing target and each diagonal crossing target within the TTC time; if the current position of a traversing target or a diagonal crossing target is within the first current conduit, and the predicted lateral position is within the first predicted conduit, then the traversing target or diagonal crossing target is determined to be a valid target; where: The constructed first current pipe has the following shape: from the front of the vehicle forward, it consists of a gradually widening section, a constant width section, a gradually narrowing section, and a final section; the width of the gradually widening section gradually increases from one end at the front of the vehicle to the other end, and the width w0 of the gradually widening section at the front of the vehicle is 0.5W to 1.5W. W represents the width of the vehicle. The width w1 at the other end of the widening section is: w1 = V_lat * TTC + offset1, where V_lat is the lateral movement speed of the target, TTC is the collision time of the target, and offset1 is a preset compensation value, offset1 ≥ 0. The width of the constant width section is the same as the width at the other end of the widening section, and the width of the constant width section remains unchanged. The width of the narrowing section gradually narrows from one end connecting to the constant width section to the other end, and the width w3 at the other end of the narrowing section is: w3 = 0.6W ~ W. The width of the final section is the same as the width at the other end of the narrowing section, and the width of the final section remains unchanged. The distance d1 between the other end of the widening section and the front of the vehicle is set as: d1 = x - offset2, and the distance d2 between the one end of the narrowing section and the front of the vehicle is set as: d2 = x + offset2. When filtering targets for active safety functions, x is the braking distance or warning distance; when filtering targets for comfort driving functions, x is the braking distance. offset2 is a preset compensation value, offset2>0; the distance d3 between the other end of the narrowing section and the front of the car is set as: d3=x+ offset2+k1*Vr, Vr is the longitudinal relative speed between the target and the car, k1 is a preset empirical value, k1=0.2 seconds~1 second, and the distance between the other end of the final section and the front of the car is set to infinity. The constructed first predictive pipeline has the following shape: from the front of the vehicle forward, it consists of a tapering section and a constant-width section; the width of the tapering section gradually narrows from one end at the front of the vehicle to the other end, with the width w6 at one end of the tapering section being 0.8W to 1.2W, and the width at the other end of the tapering section being equal to w7, with w7 being 0.3W to 1.1W; the width of the constant-width section is the same as the width at the other end of the tapering section, and the width of the constant-width section remains unchanged; the distance from the other end of the tapering section to the front of the vehicle is set as d3'', and the calculation formula for d3'' is the same as that for d3; the distance from the other end of the constant-width section to the front of the vehicle is set to infinity.

10. The target screening method according to claim 9, characterized in that, The active safety function is one of the following: Automatic Emergency Braking (AEB), Forward Collision Warning (FCW), and Emergency Steering Assist (ESA); The comfort driving function is one of the following: Adaptive Cruise Control (ACC), Traffic Jam Assist (TJA), and Navigation Assist (NOA).

11. The target screening method according to claim 9, characterized in that, When the active safety function is Automatic Emergency Braking (AEB), x represents the braking distance; when the active safety function is Forward Collision Warning (FCW) or Emergency Steering Assist (ESA), x represents the warning distance.

12. The target screening method according to claim 9 or 11, characterized in that, The formula for calculating the braking distance x is: v target Let v be the target's velocity relative to the ground. ego denoted as the vehicle speed, and 'a' as the deceleration determined by the active safety function or the comfort driving function.

13. The target screening method according to claim 9, characterized in that, offset2 equals k2*Vr, where k2 equals 0.1 seconds to 0.5 seconds.

14. A target screening device for intelligent driving, comprising at least one processor and a memory communicatively connected to said at least one processor; the memory storing instructions executable by said at least one processor; characterized in that, The at least one processor is used to execute the instructions to implement the method as described in any one of claims 9 to 13.

15. A vehicle, characterized in that, Includes the target screening device as described in claim 14.

Images