Decision control method and device for traffic light intersection, equipment and medium

Abstract

The invention relates to the field of automatic driving, and provides a traffic light intersection decision control method, device and equipment and a medium, and the method comprises the steps: enabling a decision state machine to jump to a stop line approaching state when the real-time distance between a vehicle and a stop line at a front intersection is smaller than or equal to a preset distance threshold value; in the state of approaching the stop line, traffic decision judgment is carried out based on the currently sensed green flashing / yellow light state, and the current state of the state machine is determined based on the obtained traffic decision and / or the traffic light state; when the vehicle passes through the stop line of the front intersection and drives into the front intersection, the state machine jumps from a stop line approaching state to an intersection passing state; and in the intersection passing state, executing a low-speed intersection passing decision based on the traffic light state and the position and speed of the vehicle until the vehicle leaves the front intersection. According to the method, the device, the equipment and the medium provided by the invention, key technical support is provided for reliable landing of a high-order automatic driving function.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of automatic driving, in particular to a decision control method and device for a red light intersection, equipment and a medium. BACKGROUND

[0002] With the rapid development of intelligent driving technology, the Park-to-Park (DTD) function has become a core high-frequency scenario in the field of vehicle automatic driving. This function aims to realize the autonomous driving of vehicles from the starting parking space to the target parking space, and its technical process covers key links such as automatic out-of-warehouse, public road path driving, and automatic warehousing.

[0003] However, when the driving path contains a complex scenario composed of a crossroads and a red light, the existing technical solutions have the following three bottlenecks: First, the precise perception of the red light state and the adaptability of the decision are insufficient. The traditional scheme has rigid disposal logic for uncertain states such as green flashing, yellow light, and signal obstruction, and often adopts a conservative strategy of sudden braking or parking to wait, which cannot make comprehensive judgments like human drivers, resulting in frequent function interruptions and low through efficiency. Second, the vehicle goes through multiple stages such as "lane cruise, approach stop line, and internal traffic of intersection" from approaching the intersection to driving away from the intersection, and the risk factors and decision goals of each stage are completely different. The existing control model of intersection traffic cannot cover the complex scenario with continuous changes. Third, when the vehicle has driven past the stop line but cannot pass through in time due to intersection congestion, the existing system often cannot make a human-like decision on whether to park and wait safely in the intersection, which easily leads to the vehicle being stranded in the intersection, interfering with horizontal traffic, and having serious safety hazards.

[0004] Therefore, there is an urgent need in the art for an intersection traffic control method that can deeply integrate red light state perception and have multi-scene adaptive and human-like decision-making capabilities. SUMMARY

[0005] The present application provides a decision control method, device, equipment and medium for a red light intersection, to solve the defects in the prior art that inaccurate red light calculation leads to poor function experience, rigid control logic for complex intersection scenarios, and inability to make human-like intelligent decisions for special scenarios, and to realize high-precision perception, multi-scene adaptive control and human-like intelligent decision-making of the Park-to-Park (DTD) function in complex scenarios containing crossroads and red lights, significantly improving the through efficiency, safety and ride comfort of the automatic driving system.

[0006] The present application provides a decision control method for a red light intersection, comprising: When the real-time distance between the ego vehicle and the front intersection stop line is less than or equal to a preset distance threshold, jump the decision state machine to the approach stop line state; In the approaching stop line state, a passing decision is determined based on a current perceived green flash / yellow light state, and a current state of the state machine is determined based on the passing decision and / or the red light state, the current state including: the approaching stop line state, a red light stop state, and a light state shielding state; When the ego vehicle passes the front intersection stop line and enters the inside of the front intersection, the state machine is jumped from the approaching stop line state to an intersection passing state; In the intersection passing state, a low-speed intersection passing decision is executed based on a red light state, a position and a speed of the ego vehicle, until the ego vehicle drives away from the front intersection.

[0007] According to the decision control method of the red light intersection provided by the present application, the current state of the state machine is determined based on the passing decision and / or the red light state, including: If the red light state is green light, off light, yellow light flashing, green flash / yellow light, and the passing decision is passing, it is determined that the current state is to maintain the approaching stop line state, and the vehicle is controlled to continue driving; If the red light state is red light, green flash / yellow light, and the passing decision is stopping, it is determined that the current state is the red light stop state, and the vehicle is controlled to perform a stop operation; If no red light is perceived in the current frame, it is determined that the current state is the light state shielding state, and a passing decision is determined based on the red light information or the same direction traffic information of the last frame.

[0008] According to the decision control method of the red light intersection provided by the present application, the passing decision is determined based on the current perceived green flash state, including: A green flash passable duration is calculated based on a green flash setting duration, a longitudinal target congestion loss duration, and a green flash duration, and a green flash passable distance is determined based on the green flash passable duration; A comfortable stop required distance is calculated based on a current acceleration of the ego vehicle, a maximum allowed acceleration, and a maximum allowed jerk; When the green flash passable distance is greater than the distance of the ego vehicle from the stop line, and the comfortable stop required distance is greater than the distance of the ego vehicle from the stop line, it is determined that the passing decision is passing.

[0009] According to the decision control method of the red light intersection provided by the present application, the passing decision is determined based on the same direction traffic information, including: In the case that no red light is perceived in continuous multiple frames, a target vehicle driving in the same direction as the ego vehicle is screened; A deceleration required for the target vehicle to stop is calculated based on a driving speed of the target vehicle and a distance to the intersection stop line; When the deceleration is greater than a comfort threshold, it is determined that the target vehicle intends to pass, and the passing decision of the ego vehicle is determined.

[0010] The decision control method for a traffic light intersection provided by the present application performs a low-speed intersection passing decision based on a traffic light state, a vehicle position and a speed, and comprises the following steps. When it is perceived that the traffic light is red or yellow, the vehicle has entered the intersection by a distance less than a preset proportion threshold of the total length of the intersection, and one of the following speed conditions is met, the vehicle is controlled to perform a stopping operation: The speed of the vehicle is lower than a first speed threshold; the speed of the vehicle is lower than a second speed threshold, and the sum of a planned travel distance within a preset time period in the future and a distance to pass through the intersection is less than a preset short distance threshold; the speed of the vehicle is lower than the second speed threshold, and the number of target objects followed longitudinally exceeds a preset number threshold; When it is perceived again that the traffic light is green, the vehicle is controlled to continue driving.

[0011] The decision control method for a traffic light intersection provided by the present application further comprises the following steps. When a black light is detected at the intersection, it is determined whether it is a true black light, and if a colored signal light is continuously detected within 5 seconds, it is determined as a false black light; A passing decision is made for the true black light; A decision logic corresponding to the colored signal light is used for the false black light.

[0012] The decision control method for a traffic light intersection provided by the present application further comprises the following steps. When a straight circular light and a left-turn arrow light exist at the intersection at the same time, and the left-turn arrow light is a black light, a passing decision is made based on the color of the straight circular light.

[0013] The present application also provides a decision control device for a traffic light intersection, which comprises the following steps. A first state jump unit is configured to jump a decision state machine to a stop line approach state when a real-time distance between a vehicle and a front intersection stop line is less than or equal to a preset distance threshold. A green flash passing decision unit is configured to make a passing decision based on a currently perceived green flash / yellow light state in the stop line approach state, and determine a current state of the state machine based on the obtained passing decision and / or traffic light state, wherein the current state comprises a stop line approach state, a red light stopping state and a light state occlusion state. A second state jump unit is configured to jump the state machine from the stop line approach state to an intersection passing state when the vehicle enters the front intersection inside the intersection stop line. An intersection passing decision unit is configured to perform a low-speed intersection passing decision based on a traffic light state, a vehicle position and a speed in the intersection passing state until the vehicle drives away from the front intersection.

[0014] The application further provides an electronic device, including a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor implements the decision control method of the traffic light intersection according to any one of the above when executing the computer program.

[0015] The application further provides a non-transitory computer-readable storage medium having a computer program stored thereon, wherein the computer program, when executed by a processor, implements the decision control method of the traffic light intersection according to any one of the above.

[0016] The application further provides a computer program product, including a computer program, wherein the computer program, when executed by a processor, implements the decision control method of the traffic light intersection according to any one of the above.

[0017] The application provides a decision control method, device, equipment and medium of a traffic light intersection, which realizes the prescient prediction and smooth mode switching of the intersection scene by triggering the state machine to jump based on the real-time distance. The vehicle enters the approaching stop line state in advance when it is still far away from the intersection, which reserves sufficient time and space margin for subsequent complex decision and execution, avoids the passive situation of responding in a hurry when the vehicle approaches the intersection from the source, and lays a time sequence foundation for efficient and safe passing.

[0018] The passing decision judgment based on the green flash / yellow light and the current state (approaching, parking, and occlusion) are introduced in the approaching stop line state, which solves the decision blind area of the traditional rule of non-red or green. In the face of ambiguous signals such as green flash and yellow light, the vehicle can make a comprehensive quantitative calculation like an experienced driver. At the same time, in the face of signal light occlusion, it can seamlessly switch to the light state occlusion state, ensuring the continuity of decision-making under the uncertainty of perception, and significantly improving the smoothness and intelligence level of passing.

[0019] By jumping to the intersection passing state after entering the intersection and executing the low-speed intersection passing decision, the fine management and control of the high-risk area in the intersection are realized. The vehicle can be effectively prevented from being in a dilemma due to signal mutation or severe congestion after passing the stop line. The vehicle can actively judge and park at a safe position in the intersection to wait, avoiding blocking the horizontal traffic, so as to ensure the overall traffic order and safety in extreme cases, and embodies the high-level human-like decision-making ability.

[0020] In summary, the progressive and full-process decision control mechanism from “long-distance perception warning” to “near-intersection intelligent judgment” to “interior-intersection active fault tolerance” converts the discrete intersection scene into a continuous and controllable state migration process, and finally achieves the three goals of improving passing efficiency, ensuring driving safety, and optimizing riding experience in the complex and variable real intersection environment, which provides key technical support for the reliable landing of high-level automatic driving functions. BRIEF DESCRIPTION OF DRAWINGS

[0021] In order to more clearly illustrate the technical solutions in the present application or the prior art, the following will briefly introduce the drawings needed to be used in the embodiments or prior art description. Obviously, the drawings in the following description are some embodiments of the present application, and other drawings can also be obtained by those skilled in the art without creative labor.

[0022] Figure 1 is a flowchart of the decision control method of the traffic light intersection provided by the present application.

[0023] Figure 2 is a logic diagram of the green flash passing decision provided by the present application.

[0024] Figure 3 is a passing decision flowchart in the occlusion state provided by the present application.

[0025] Figure 4 is a state machine jump logic diagram provided by the present application.

[0026] Figure 5 is a state machine and intersection position relationship diagram provided by the present application.

[0027] Figure 6 is a structural diagram of the decision control device of the traffic light intersection provided by the present application.

[0028] Figure 7 is a structural diagram of the electronic device provided by the present application. DETAILED DESCRIPTION

[0029] In order to make the objects, technical solutions and advantages of the present application clearer, the following will combine the drawings in the present application to clearly and completely describe the technical solutions in the present application. Obviously, the described embodiments are some embodiments of the present application, not all embodiments. Based on the embodiments in the present application, all other embodiments obtained by those skilled in the art without creative labor are within the protection scope of the present application.

[0030] The embodiments of the present application can be applied to scenes requiring complex intersection intelligent decision and control, such as urban road automatic driving, intersection passing of intelligent networked vehicles controlled by traffic lights, and car-to-car (DTD) full-process autonomous driving scenes. The execution subject of the present method can be a terminal device, a vehicle-mounted computing platform, an automatic driving domain controller, an edge server, a cloud server cluster, or a specially designed vehicle-road cooperation roadside unit electronic device. It can also be a traffic light intersection decision control device arranged in the electronic device. The device can be realized by software, hardware, or a combination of both.

[0031] In the description of the embodiments of the present application, it should be understood that the terms "first", "second" are only used for descriptive purposes, and cannot be understood as indicating or implying relative importance or implicitly indicating the number of indicated technical features. Therefore, the features defined with "first", "second" can explicitly or implicitly include one or more of the features. In the description of the embodiments of the present application, the meaning of "multiple" is two or more, unless otherwise specifically limited.

[0032] Figure 1 is a flowchart of the decision control method of the traffic light intersection provided by the present application, as shown in Figure 1 The method comprises steps 110 to 140.

[0033] Step 110, when the real-time distance between the ego vehicle and the front intersection stop line is less than or equal to the preset distance threshold, jump to the approaching stop line state.

[0034] Specifically, the real-time distance between the ego vehicle and the front intersection stop line can be calculated by global positioning system (GPS), laser radar, camera and high-precision map fusion positioning technology, etc. The straight-line distance from the ego vehicle (e.g. vehicle center of mass) to the front intersection stop line can be calculated in real time. The distance is a dynamically updated value. The preset distance threshold is a set distance value, which can be determined according to the vehicle performance, control system reaction time and safety margin, and the purpose is to ensure that the vehicle has enough distance and time to operate, such as deceleration, parking or path planning. The preset distance threshold may be, for example, 90 meters.

[0035] The decision state machine is a mathematical model or logic framework used to describe the behavior of the system in different states, as well as the rules for switching between states. The approaching stop line state is a specific state in the decision state machine. Entering this state means that the system has officially started the processing logic for the traffic light intersection, such as starting to detect the signal light more frequently, evaluating the right of way, preparing to brake, etc.

[0036] After the vehicle enters the autonomous driving state, the traffic light decision state machine is initialized to the initial state (INIT). After accumulating one frame of valid traffic light information, the state machine is jumped to the lane keeping state (LANE_KEEPING).

[0037] The system acquires geometric information of the upcoming intersection through a high-precision map, especially the precise coordinates of the stop line. Simultaneously, it utilizes a combined positioning system (GPS + IMU + wheel speedometer, etc.) to obtain the vehicle's own precise coordinates in real time. The control unit calculates the Euclidean distance between the vehicle's coordinates and the stop line coordinates in real time. In each control cycle (e.g., 10 milliseconds), the calculated real-time distance is compared with a preset threshold (90 meters). If the real-time distance is greater than the first preset distance threshold, the lane-keeping state (LANE_KEEPING) is maintained; once the condition that the real-time distance is less than or equal to the first preset distance threshold is met, the system immediately switches the current state from lane-keeping state to approach-stop-line state (APPROACH_STOPLINE).

[0038] This step allows for early detection of traffic light intersections and proactive switching to the appropriate response mode, providing ample time and space for subsequent refined decision-making. It avoids hasty decisions when vehicles are almost at the intersection, thus ensuring both safety and improving efficiency.

[0039] Step 120: When approaching the stop line, make a passage decision based on the currently perceived green / yellow light status, and determine the current state of the state machine based on the obtained passage decision and / or traffic light status. The current state includes: approaching the stop line, red light stop, and light obstruction.

[0040] Specifically, the purpose of this step is to make a real-time decision to proceed or stop when a vehicle approaches the stop line at a critical stage, based on the intelligent judgment of complex and dynamic traffic light states, especially transitional signals such as flashing green and yellow lights, and to precisely switch the control state of the system accordingly.

[0041] The flashing green / yellow light status refers to the transitional signal between a stable green light and a stable red light during the traffic light cycle. A flashing green light indicates that the green light period is about to end, while a yellow light reminds vehicles to be cautious and prepare to stop. Judging these statuses is both the challenge and the key to making traffic decisions.

[0042] Making a passage decision based on the currently perceived flashing green / yellow light status is a comprehensive analysis process. The system not only identifies the color of the traffic light, but also combines multiple factors such as the duration of the flashing green light, the vehicle's speed, the distance to the stop line, and the behavior of vehicles ahead to calculate whether to accelerate through or decelerate and stop.

[0043] In some embodiments, making a passage decision based on the currently perceived green flashing state includes: Based on the green flash setting duration, the longitudinal target congestion loss duration, and the green flash duration, calculate the green flash passable duration, and determine the green flash passable distance based on the green flash passable duration; Calculate the distance required for comfortable braking based on the vehicle's current acceleration, maximum allowable acceleration, and maximum allowable jerk. When the green flashing light allows passage at a distance greater than the distance between the vehicle and the stop line, and the distance required for a comfortable stop is also greater than the distance between the vehicle and the stop line, a passage decision is made.

[0044] Specifically, Figure 2 This is a logical diagram of the green flashing traffic decision-making system provided by this invention. The green flashing duration refers to the preset total duration of the green flashing signal in the traffic light system, for example, 2.1 seconds, representing the theoretical maximum travel time window. The longitudinal target congestion loss time refers to the reduction in effective travel time caused by vehicles ahead (longitudinal targets). For example, for each additional vehicle ahead, the travel time decreases by 0.1 seconds. The green flashing duration refers to the time elapsed from the start of the green flashing signal to the current moment, and is a real-time changing measurement.

[0045] The green flashing passable time refers to the remaining effective passage time available for use. The formula for calculating the green flashing passable time is: Green flashing set duration - Longitudinal target congestion loss time - Green flashing duration. The green flashing passable distance refers to the distance a vehicle is expected to travel based on its current speed and control strategy within the remaining green flashing passable time.

[0046] The distance required for a comfortable braking stop refers to the shortest distance a vehicle needs to smoothly decelerate from its current speed to a stop at a rate of deceleration that is comfortable for the human body. The calculation of this distance takes into account both safety and passenger experience, avoiding sudden braking.

[0047] In practice, the system first obtains the set duration of the green flashing light through vehicle-to-infrastructure communication or visual recognition, and then uses sensors to detect the number of vehicles ahead and the duration of the green flashing light to calculate the dynamic green flashing passability duration in real time. Subsequently, based on the current vehicle speed and this duration, the passability distance under the green flashing light is estimated using kinematic formulas.

[0048] At the same time, another calculation path is executed in parallel: read the vehicle's current acceleration and, based on the preset maximum allowable acceleration and maximum allowable jerk (Jerk), calculate the distance required for a comfortable and smooth stop.

[0049] Finally, these two calculated distances are compared in real time with the distance from the vehicle to the stop line. A passage decision is made only when both of the following conditions are met: First, the distance that can be passed under the flashing green light is greater than the distance between the vehicle and the stop line (meaning the vehicle can cross the stop line before the green light turns off); second, the distance required for a comfortable stop is greater than the distance between the vehicle and the stop line (meaning it's too late to stop comfortably at the stop line). When the deceleration required for the vehicle to stop is very large, exceeding the comfort threshold based on the vehicle's speed, a passage decision will be made. Through this dual-condition judgment, the decision-making logic of a human driver—"if you can cross, then cross; if you can't, then stop calmly"—is cleverly simulated.

[0050] The embodiments of the present invention introduce computation-based traffic decision-making, so that the system's behavior in scenarios such as flashing green lights is no longer a rigid "go / stop" dichotomy, but a human-like decision that integrates safety, efficiency and comfort, significantly improving the driving experience.

[0051] Setting a more conservative traffic decision while ensuring safety can improve traffic efficiency and avoid excessive braking in some scenarios. At the same time, a more conservative green flashing decision avoids a series of problems within the intersection caused by the cross lane turning green after entering the intersection.

[0052] Based on this, the current state of the state machine is determined according to the obtained traffic decision and / or traffic light status. In some embodiments, this specifically includes: If the traffic light status is green, off, flashing yellow, or flashing green / yellow and the passage decision is to proceed, then the current state is determined to be maintaining proximity to the stop line, and the vehicle is controlled to continue moving. If the traffic light status is red, flashing green / yellow and the passage decision is to stop, then the current status is determined to be red light stop status, and the vehicle is controlled to perform a stop operation; If no traffic light is detected in the current frame, the current state is determined to be a light occlusion state, and a passage decision is made based on the traffic light information or same-direction traffic flow information of the previous frame.

[0053] Specifically, this embodiment describes the core decision-making logic of the state machine when approaching the stop line. Based on real-time perception results, it determines whether the vehicle should continue approaching, stop, or activate the backup plan. The current state is during the vehicle's approach to the intersection, and the state machine may be in one of three key states: Approaching the stop line: The vehicle is approaching the intersection normally, and the system continues to perceive and make decisions. Stopping at a red light: The system has made a stopping decision, and the vehicle enters the deceleration and stopping process. Traffic light obstruction: The traffic light signal source is malfunctioning (e.g., obstructed by the vehicle in front), and the backup decision mode is activated.

[0054] In each control cycle, two key inputs are acquired: the current traffic light status output by the sensing module, and the passage decision result calculated by the decision module for scenarios such as flashing green / yellow lights. These inputs are matched with preset rules to drive the state machine transition. If the status is green light, off light, or flashing yellow light, or if the status is flashing green / yellow light and the passage decision is "pass", the state machine remains in the near-stop line state and controls the vehicle to continue moving.

[0055] If the light is red, or if the light is flashing green / yellow but the decision to proceed is "stop", the state machine immediately jumps to the red light stop state and triggers the braking system to perform a comfortable stop.

[0056] If the perception module reports "no traffic light detected" (i.e., information in the current frame is missing), the state machine transitions to the light occlusion state. Subsequently, the occlusion passage decision algorithm is invoked to make a comprehensive judgment based on the traffic light status of the previous frame or the behavioral intentions of vehicles traveling in the same direction, thereby maintaining decision-making capability.

[0057] After the state machine transitions, the corresponding state will be bound to specific control commands, such as constant speed driving, deceleration and stopping, and following the vehicle in front, which will eventually translate into precise control of the accelerator and brake.

[0058] This invention maps specific and diverse traffic light states to a finite set of state machine states, achieving a unified processing framework for both routine (green light, red light) and edge (flashing green, obstruction) scenarios. This enables the system to handle intersection complexities with ease. The light obstruction state and backup decision-making mechanism ensure that even in the event of a common failure such as transient perception malfunction, safety decisions can still be made based on historical information or environmental context (traffic flow in the same direction). This significantly reduces the probability of functional interruption due to momentary blind spots, meeting the high reliability requirements of automotive-grade functional safety.

[0059] In some embodiments, making traffic flow decisions based on same-direction traffic information includes: If no traffic lights are detected for several consecutive frames, select target vehicles traveling in the same direction as your own vehicle. Calculate the deceleration required for the target vehicle to come to a stop based on the target vehicle's speed and the distance to the stop line at the intersection. When the deceleration is greater than the comfort threshold, the target vehicle is determined to pass, and the vehicle's passing decision is determined accordingly.

[0060] Specifically, Figure 3 This is a flowchart of the traffic decision-making process under occlusion conditions provided by the present invention. The purpose of this embodiment is to solve the decision paralysis problem faced by autonomous driving systems due to the loss of direct right-of-way information during multiple frames when traffic lights are continuously occluded. By analyzing the collective behavioral intentions of traffic flow in the same direction, an indirect, context-based basis for right-of-way judgment is provided to the system, thereby significantly improving robustness and traffic efficiency in complex urban intersection scenarios.

[0061] Continuous multi-frame refers to the situation where the autonomous driving system fails to detect a valid traffic light signal for multiple consecutive processing cycles (e.g., 10 frames per second, or 5 consecutive frames, or 0.5 seconds), indicating that the obstruction is continuous rather than a momentary interference.

[0062] Target vehicles traveling in the same direction are those traveling in the same lane or adjacent lanes in the same direction, and whose direction of travel is consistent with that of your own vehicle. Screening criteria typically include lateral position and heading angle to ensure that the target vehicle and your own vehicle are facing the same traffic light conditions.

[0063] The deceleration required to stop is used to quantify the braking intensity required for a target vehicle to come to a smooth stop before the stop line under its current condition. This value reflects the urgency of the vehicle in front. For the selected target vehicle, by obtaining its travel speed v and distance to the stop line dist_to_stopline_, the deceleration required to stop is estimated as a = v² / 2 * (dist_to_stopline_). When the deceleration required to stop is greater than the comfort threshold, the target vehicle is considered to intend to proceed, because comfortable braking is no longer sufficient to stop safely. When the deceleration required to stop is less than or equal to the comfort threshold, the target vehicle is considered to intend to stop. The comfort threshold here is a preset empirical value, representing the upper limit of comfortable braking deceleration that a human driver can accept under normal conditions, for example, 2.5 m / s². 2 When the calculated deceleration exceeds this threshold, it means that the vehicle in front would need to brake suddenly to stop, which does not conform to normal driving logic.

[0064] The intentions of all vehicles heading in the same direction can be considered, such as through a voting mechanism. If the majority of vehicles or the nearest vehicle intends to "pass," the vehicle will make a passing decision; otherwise, it will make a stopping decision.

[0065] The method provided in this invention infers right-of-way by observing the behavior of other vehicles, highly simulating the decision-making process of human drivers in similar scenarios, making autonomous driving behavior more natural and predictable. It avoids blind sudden braking or risky passage due to missing information, enabling vehicles to integrate into traffic flow and make safe and efficient decisions, reducing traffic disruptions and accident risks.

[0066] Step 130: When the vehicle passes the stop line of the intersection ahead and enters the intersection, the state machine is switched from the approach to the stop line state to the intersection passage state.

[0067] Specifically, the area inside an intersection refers to the region within the geometric boundaries of the intersection after a vehicle has crossed the stop line. This area is characterized by intersecting traffic flows, potential lack of lane markings, and the need to guard against lateral vehicles and pedestrians, making it a high-risk area for conflicts. During intersection traffic flow, the system needs to handle the unique logic within the intersection, such as obstacle avoidance, responding to emergencies, and deciding whether to stop and wait under specific conditions.

[0068] The system continuously calculates the relative position of the vehicle to the stop line by integrating high-precision maps (including precise coordinates of the stop line) and real-time vehicle positioning. Once it detects that the vehicle's center of gravity or front bumper has crossed the stop line coordinates, it immediately triggers a state transition, changing the state from the approaching stop line state to the intersection passage state. When the vehicle's front wheels cross the stop line, it means that it has physically entered the intersection area, and the right-of-way rules and risk environment change. At this time, the control state needs to be switched from the preparatory mode of approaching the intersection to the execution mode of passing within the intersection, in order to activate a dedicated control strategy for the complex scenarios inside the intersection.

[0069] Step 140: While the vehicle is in the intersection passage state, make a low-speed intersection passage decision based on the traffic light status, vehicle position and speed, until the vehicle leaves the intersection ahead.

[0070] Specifically, the purpose of this step is to safely and orderly handle special scenarios where vehicles cannot immediately pass through intersections due to unforeseen circumstances, such as changes in traffic lights or severe congestion. The design aims to prevent vehicles from becoming stuck in the middle of intersections, thereby obstructing traffic and causing safety accidents.

[0071] Low-speed intersection passage decision-making is a risk assessment and decision-making algorithm specifically designed for intersections. The decision result is not always "pass"; it may also be "safely stop and wait within the intersection." This decision simulates the judgment ability of an experienced driver within an intersection, that is, when the right of way is uncertain, choosing not to enter or safely stopping and waiting, rather than risking to rush through.

[0072] In some embodiments, making low-speed intersection passage decisions based on traffic light status, vehicle position, and speed includes: When the traffic light is detected to be red or yellow, the distance the vehicle has traveled into the intersection is less than a preset threshold proportion of the total length of the intersection, and one of the following speed conditions is met, the vehicle will be controlled to stop: The vehicle's speed is lower than the first speed threshold; the vehicle's speed is lower than the second speed threshold, and the sum of the planned travel distance and the distance to the intersection within a preset time period is less than a preset short distance threshold; the vehicle's speed is lower than the second speed threshold, and the number of targets following longitudinally exceeds a preset number threshold. When the traffic light is detected to be green again, the vehicle will continue to move.

[0073] Specifically, in scenarios where a vehicle has passed the stop line at an intersection and entered the intersection, but is unable to continue due to a sudden change in the traffic light (such as turning red or yellow) or severe congestion, in order to prevent the vehicle from being trapped in the middle of the intersection, thereby avoiding obstruction of cross traffic and causing safety accidents, and to achieve intelligent judgment similar to that of a human driver, this embodiment selects the safest location within the intersection to stop and wait until conditions permit before continuing to proceed.

[0074] The distance a vehicle has traveled within the intersection is the distance it has already covered from the stop line within the intersection. The preset percentage threshold for the total intersection length refers to the physical length of the entire intersection, such as the distance from the stop line to the opposite curb multiplied by a preset percentage, such as 25%. This threshold defines a safe zone where vehicles stopping will not significantly obstruct cross traffic.

[0075] The first speed threshold and the second speed threshold are speed cutoff values ​​used to determine whether a vehicle is in a creeping or low-speed state. For example, the first speed threshold could be 0.3 m / s, and the second speed threshold could be 1.0 m / s.

[0076] The preset short-distance threshold refers to the upper limit of the sum of the planned distance a vehicle can travel and the remaining distance required to pass through the intersection within a very short period of time (such as 3 seconds). For example, the short-distance threshold could be 10 meters, which is used to determine whether a vehicle cannot leave the intersection in time.

[0077] The preset threshold is a critical value for the number of vehicles following longitudinally, such as >4 vehicles. This value indicates severe congestion ahead and slow-moving convoy.

[0078] The decision-making process for passing through a low-speed intersection is an automated process triggered by multiple conditions. When the state machine is in the intersection passage state, the system continuously monitors conditions in three dimensions: Traffic lights: Real-time detection of whether the light is red or yellow; flashing green at long intersections.

[0079] Location: Inside the intersection and where the traffic lights are visible, meaning the traffic lights are within the camera's field of view.

[0080] Speed ​​conditions (one of the following must be met): Speed ​​condition 1: Vehicle speed < 0.3 m / s Speed ​​condition 2: Vehicle speed < 1.0 m / s and planned passable distance + distance to intersection within 3 seconds < 10 meters. Speed ​​condition 3: Vehicle speed < 1.0 m / s and number of longitudinally following targets > 4 Intersection distance condition: The distance of the vehicle through the intersection is less than 25% of the intersection length. When the intersection length * 25% is less than 7m, it is counted as 7m.

[0081] Traffic conditions: The vehicle itself makes the traffic decision.

[0082] When all the above conditions are met simultaneously (i.e., traffic light deterioration + vehicle within the safe zone at the intersection entrance + vehicle in a low-speed congestion state), the system immediately triggers a "low-speed intersection passage decision" and outputs a "stop" command. The vehicle begins to decelerate smoothly and comes to a safe stop within the intersection. After the vehicle stops, the system continues to monitor the traffic lights. When the green light is detected again, indicating that the right of way has been restored, the system immediately cancels the stop command, controls the vehicle to accelerate again, and continues to pass through the intersection.

[0083] In some embodiments, Figure 4 This is a schematic diagram of the state machine transition logic provided by the present invention. Figure 5 This is a schematic diagram illustrating the relationship between the state machine and the intersection position provided by the present invention, as shown below. Figure 4 and Figure 5 As shown, external environment perception is achieved through multi-sensor fusion technology. Vertically, a dual-loop PID + feedforward control algorithm (position loop + velocity loop) is used, while horizontally, MPC control logic is employed to jointly complete the control of the entire DTD function. The state machine transition logic is as follows: ① Each time the system enters autonomous driving mode, it accumulates one frame of traffic light information, and the initial INIT state will switch to the lane keeping LANE_KEPPING state.

[0084] ②Reference Figure 5 Looking from bottom to top, when the vehicle is more than 90 meters from the stop line at the intersection, the state machine is in the LANE_KEPPING state. When the distance to the stop line is less than 90 meters, the state machine enters the APPROACH_STOPLINE state.

[0085] ③ In APPROACH_STOPLINE state, when the traffic light is green, or the green light is flashing / yellow light indicates passage, or the light is off, or the yellow light is flashing, the vehicle continues to drive while maintaining APPROACH_STOPLINE state; When the traffic light is red, or the decision to stop is a flashing green / yellow light, the state machine will enter the RED_LIGHT_STOP state. When the traffic light is blocked, the state machine will enter the COVER_LIGHT state.

[0086] ④ In the RED_LIGHT_STOP state, the state machine will make a stopping decision. When the traffic light turns green or OFF, it will re-enter the APPROACH_STOPLINE state.

[0087] ⑤ In the COVER_LIGHT state, when a green light, a flashing yellow light, or a light that is off are seen, or when a flashing green light is seen and a decision to proceed is made, the state machine re-enters the APPROACH_STOPLINE state. When a red light is seen, or when a flashing green light is seen and a decision to stop is made, the state machine will enter the RED_LIGHT_STOP state. If no light is seen for several consecutive frames, the occlusion decision will be made by combining the state of the traffic light seen in the last frame, the vehicle's speed, and the distance from the intersection.

[0088] ⑥ In the APPROACH_STOPLINE state, when the vehicle passes the stop line at the intersection and enters the intersection, the state machine will enter the CROSSING state.

[0089] ⑦ In the CROSSING state at an intersection, if the traffic light is red or yellow and the conditions for a low-speed intersection passage decision are met, the vehicle will make a stop decision. It will then make a passage decision again upon seeing the green light again. After leaving the intersection, the state machine will transition from the CROSSING state to the LANCE_KEPPING state.

[0090] In other embodiments, the decision control method further includes: When a black light is detected at an intersection, it is determined whether it is a true black light. If a colored traffic light is continuously detected within 5 seconds, it is determined to be a false black light. For truly dark lights, a pass-through decision is made; The decision logic for a false black light is to use the corresponding color signal light.

[0091] Specifically, to address the decision-making dilemma faced by autonomous driving systems when encountering black light signals at intersections, this embodiment employs a rigorous true / false signal discrimination logic. This prevents genuine traffic light malfunctions (true black lights) from being misjudged as red lights, leading to incorrect stopping and thus improving traffic efficiency. Furthermore, it prevents sensor misdetections (false black lights, actually red / green lights) from being misjudged as permissible, thus avoiding traffic accidents and ensuring safety. Ultimately, this ensures that the system can make human-like, reasonable, and reliable traffic decisions even in black light scenarios.

[0092] Here, "black lights" at an intersection refers to a state where all traffic lights at the intersection are off or not lit, as detected by sensors. This is a raw perception result, and its accuracy is unknown. "True black lights" means that the traffic light control system at the intersection has indeed malfunctioned, and all lights are off. "False black lights" means that the traffic lights at the intersection are actually working normally (e.g., red or green lights are on), but due to sensor misidentification, brief obstruction, or light interference, the system incorrectly identifies them as black lights. This is a type of perception error.

[0093] In practice, the system initially judges the black light result output by the perception module as a genuine black light and prepares to make a passage decision. Perception continues for the next 5 seconds. If, at any moment within this time window, any colored traffic light (such as red, green, or yellow) is clearly detected at the intersection, a false black light determination is immediately triggered.

[0094] If the light is determined to be truly dark, a passage decision is adopted, treating the intersection as if it were in a state of light outage or malfunction, and controlling vehicles to pass through only when it is safe to do so.

[0095] If the light is determined to be a false black light, the black light signal will be ignored, and the system will switch to the decision logic corresponding to the identified colored traffic light (such as the red light) (such as the RED_LIGHT_STOP state) to control the vehicle to stop and wait.

[0096] It should be noted that each time a vehicle passes the stop line at an intersection, the flag for determining whether the light is off will be reset to its default value, preparing for the next time a vehicle enters the intersection.

[0097] This invention, through an intelligent diagnostic mechanism of "true and false black lights," effectively overcomes the uncertainty of sensor perception, readily handling a common type of perception error and ensuring functional continuity and reliability. Proceeding through true black lights avoids unnecessary stopping delays; for false black lights, it adopts the actual traffic light rules, eliminating the risk of running red lights. Thus, it maximizes traffic efficiency while ensuring safety. Furthermore, this solution goes beyond simple "perception-action" mapping, introducing continuous observation and logical reasoning over time. Like an experienced driver, it can verify suspicious traffic signals before making decisions, representing a typical manifestation of human-like thinking.

[0098] In some embodiments, the method further includes: When an intersection has both a straight-ahead round light and a left-turn arrow light, and the left-turn arrow light is black, the color of the straight-ahead round light is used to make the passage decision.

[0099] Specifically, this embodiment addresses the challenge of determining the right-of-way for left turns at intersections with a combination of straight-ahead round lights and left-turn arrow lights, when the left-turn arrow light is malfunctioning and displays as black.

[0100] A straight-ahead round light is a circular traffic signal light that indicates the direction of going straight. Its color (red, yellow, green) controls the right-of-way for vehicles going straight. A left-turn arrow light is an arrow-shaped traffic signal light that specifically indicates the direction of turning left. Its color independently controls the right-of-way for vehicles turning left. A black left-turn arrow light means that the left-turn arrow light is off or not lit, usually indicating that the signal light group may be malfunctioning, or that the dedicated left-turn phase is not in use during that time period.

[0101] Vehicle perception systems, such as cameras, first identify the traffic light groups at the intersection and accurately classify them as "straight-ahead round lights" and "left-turn arrow lights." Then, they detect the status of the left-turn arrow light. If its color is detected as valid red, yellow, or green, the system makes a decision according to normal left-turn signal logic. If it is detected as "black" (i.e., no valid color can be identified), the system ignores the left-turn arrow light's status and binds its passage decision to the straight-ahead round light. The system reads the real-time color of the straight-ahead round light and directly applies the corresponding decision rule, such as green / off / flashing yellow - pass, red - stop, to left-turn control. Based on the mapped decision, the system controls the vehicle to perform the corresponding action—either proceed or stop.

[0102] This invention effectively addresses common fault scenarios such as left-turn signal light failure, preventing system crashes due to incomplete perception information and ensuring the continuity and availability of autonomous driving functions. It avoids situations where vehicles are stuck at intersections for extended periods due to the inability to obtain a clear left-turn signal, thus helping to maintain smooth traffic flow and improving overall traffic efficiency.

[0103] The decision control device for traffic light intersections provided by the present invention will be described below. The decision control device for traffic light intersections described below can be referred to in correspondence with the decision control method for traffic light intersections described above.

[0104] Based on the above embodiments, Figure 6 This is a schematic diagram of the decision control device for traffic light intersections provided by the present invention, as shown below. Figure 6 As shown, the device includes: The first state transition unit 610 is used to transition the decision state machine to the near-stop line state when the real-time distance between the vehicle and the stop line of the intersection ahead is less than or equal to a preset distance threshold. The green flashing passage decision unit 620 is used to make a passage decision judgment based on the currently perceived green flashing / yellow light state when approaching the stop line state, and to determine the current state of the state machine based on the obtained passage decision and / or traffic light state. The current state includes: approaching the stop line state, red light stop state, and light state obstruction state. The second state transition unit 630 is used to transition the state machine from the approach to the stop line state to the intersection passage state when the vehicle passes the stop line of the intersection and enters the intersection ahead. The intersection passage decision unit 640 is used to make a low-speed intersection passage decision based on the traffic light status, vehicle position and speed during the intersection passage state, until the vehicle leaves the intersection ahead.

[0105] It is understood that the technical features and the technical effects achieved by the embodiments of the present invention can be referred to the descriptions of the above embodiments, and will not be repeated here.

[0106] Figure 7 An example is a schematic diagram of the physical structure of an electronic device, such as... Figure 7 As shown, the electronic device may include: a processor 710, a communications interface 720, a memory 730, and a communications bus 740, wherein the processor 710, the communications interface 720, and the memory 730 communicate with each other through the communications bus 740. The processor 710 can call logic instructions in the memory 730 to execute a decision control method for traffic light intersections. This method includes: when the real-time distance between the vehicle and the stop line of the intersection ahead is less than or equal to a preset distance threshold, switching the decision state machine to a near-stop line state; in the near-stop line state, making a passage decision based on the currently perceived flashing green / yellow light status, and determining the current state of the state machine based on the obtained passage decision and / or traffic light status, the current state including: near-stop line state, red light stop state, and light obstruction state; when the vehicle passes the stop line of the intersection ahead and enters the intersection, switching the state machine from the near-stop line state to a passage state; in the passage state, executing a low-speed intersection passage decision based on the traffic light status, the vehicle's position, and speed until the vehicle leaves the intersection.

[0107] Furthermore, the logical instructions in the aforementioned memory 730 can be implemented as software functional units and, when sold or used as independent products, can be stored in a computer-readable storage medium. Based on this understanding, the technical solution of the present invention, essentially, or the part that contributes to the prior art, or a part of the technical solution, can be embodied in the form of a software product. This computer software product is stored in a storage medium and includes several instructions to cause a computer device (which may be a personal computer, server, or network device, etc.) to execute all or part of the steps of the methods described in the various embodiments of the present invention. The aforementioned storage medium includes various media capable of storing program code, such as USB flash drives, portable hard drives, read-only memory (ROM), random access memory (RAM), magnetic disks, or optical disks.

[0108] On the other hand, the present invention also provides a computer program product, which includes a computer program that can be stored on a non-transitory computer-readable storage medium. When the computer program is executed by a processor, the computer can execute the decision control method for traffic light intersections provided by the above methods. The method includes: when the real-time distance between the vehicle and the stop line of the intersection ahead is less than or equal to a preset distance threshold, switching the decision state machine to a near-stop line state; in the near-stop line state, making a passage decision based on the currently perceived green / yellow light status, and determining the current state of the state machine based on the obtained passage decision and / or traffic light status, the current state including: near-stop line state, red light stop state, and light obstruction state; when the vehicle passes the stop line of the intersection ahead and enters the intersection ahead, switching the state machine from the near-stop line state to the intersection passage state; in the intersection passage state, performing a low-speed intersection passage decision based on the traffic light status, the vehicle's position, and speed, until the vehicle leaves the intersection ahead.

[0109] In another aspect, the present invention also provides a non-transitory computer-readable storage medium storing a computer program thereon. When executed by a processor, the computer program implements a decision control method for traffic light intersections provided by the above methods. The method includes: when the real-time distance between the vehicle and the stop line of the intersection ahead is less than or equal to a preset distance threshold, switching the decision state machine to a near-stop line state; in the near-stop line state, making a passage decision based on the currently perceived flashing green / yellow light state, and determining the current state of the state machine based on the obtained passage decision and / or traffic light state, the current state including: near-stop line state, red light stop state, and light obstruction state; when the vehicle passes the stop line of the intersection ahead and enters the intersection ahead, switching the state machine from the near-stop line state to the intersection passage state; in the intersection passage state, performing a low-speed intersection passage decision based on the traffic light state, the vehicle's position, and speed, until the vehicle leaves the intersection ahead.

[0110] The device embodiments described above are merely illustrative. The units described as separate components may or may not be physically separate. The components shown as units may or may not be physical units; that is, they may be located in one place or distributed across multiple network units. Some or all of the modules can be selected to achieve the purpose of this embodiment according to actual needs. Those skilled in the art can understand and implement this without any creative effort.

[0111] Through the above description of the embodiments, those skilled in the art can clearly understand that each embodiment can be implemented by means of software plus necessary general-purpose hardware platforms, and of course, it can also be implemented by hardware. Based on this understanding, the above technical solutions, in essence or the part that contributes to the prior art, can be embodied in the form of a software product. This computer software product can be stored in a computer-readable storage medium, such as ROM / RAM, magnetic disk, optical disk, etc., and includes several instructions to cause a computer device (which may be a personal computer, server, or network device, etc.) to execute the methods described in the various embodiments or some parts of the embodiments.

[0112] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention, and not to limit them; although the present invention has been described in detail with reference to the foregoing embodiments, those skilled in the art should understand that modifications can still be made to the technical solutions described in the foregoing embodiments, or equivalent substitutions can be made to some of the technical features; and these modifications or substitutions do not cause the essence of the corresponding technical solutions to deviate from the spirit and scope of the technical solutions of the embodiments of the present invention.

Claims

1. A decision-making and control method for traffic light intersections, characterized in that, include: When the real-time distance between the vehicle and the stop line at the intersection ahead is less than or equal to a preset distance threshold, the decision state machine will switch to the approaching stop line state. In the state of approaching the stop line, a passage decision is made based on the currently perceived green flashing / yellow light state, and the current state of the state machine is determined based on the obtained passage decision and / or traffic light state. The current state includes: approaching the stop line state, red light stop state, and light state obstruction state. When the vehicle passes the stop line at the intersection ahead and enters the intersection, the state machine will switch from the approach to the stop line state to the intersection passage state. When the vehicle is in the intersection passage state, it makes a low-speed intersection passage decision based on the traffic light status, vehicle position and speed, until the vehicle leaves the intersection ahead.

2. The decision-making and control method for traffic light intersections according to claim 1, characterized in that, Determining the current state of the state machine based on the passage decision and / or traffic light status includes: If the traffic light status is green, off, flashing yellow, or flashing green / yellow and the passage decision is to proceed, then the current state is determined to be maintaining proximity to the stop line, and the vehicle is controlled to continue moving. If the traffic light status is red, flashing green / yellow and the passage decision is to stop, then the current status is determined to be a red light stop status, and the vehicle is controlled to perform a stop operation; If no traffic light is detected in the current frame, the current state is determined to be a light obstruction state, and a passage decision is made based on the traffic light information or same-direction traffic flow information of the previous frame.

3. The decision-making and control method for traffic light intersections according to claim 2, characterized in that, The decision to allow passage is based on the currently perceived green flashing status, including: Based on the green flashing setting duration, the longitudinal target congestion loss duration, and the green flashing duration, the green flashing passable duration is calculated, and the green flashing passable distance is determined based on the green flashing passable duration. Calculate the distance required for comfortable braking based on the vehicle's current acceleration, maximum allowable acceleration, and maximum allowable jerk. When the green flashing light allows passage at a distance greater than the distance between the vehicle and the stop line, and the distance required for a comfortable stop is also greater than the distance between the vehicle and the stop line, a passage decision is made.

4. The decision-making and control method for traffic light intersections according to claim 2, characterized in that, The decision-making process based on same-direction traffic flow information includes: If no traffic lights are detected for several consecutive frames, select target vehicles traveling in the same direction as your own vehicle. Calculate the deceleration required for the target vehicle to come to a stop based on the target vehicle's speed and the distance to the stop line at the intersection. When the deceleration is greater than the comfort threshold, it is determined that the target vehicle intends to pass, and the vehicle's passing decision is determined.

5. The decision-making and control method for traffic light intersections according to claim 1, characterized in that, The method of making low-speed intersection passage decisions based on traffic light status, vehicle position, and speed includes: When the traffic light is detected to be red or yellow, the distance the vehicle has traveled into the intersection is less than a preset threshold proportion of the total length of the intersection, and one of the following speed conditions is met, the vehicle will be controlled to stop: The vehicle's speed is lower than the first speed threshold; the vehicle's speed is lower than the second speed threshold, and the sum of the planned travel distance and the distance to the intersection within a preset time period is less than a preset short distance threshold; the vehicle's speed is lower than the second speed threshold, and the number of targets following longitudinally exceeds a preset number threshold. When the traffic light is detected to be green again, the vehicle will continue to move.

6. The decision-making and control method for traffic light intersections according to claim 1, characterized in that, The method further includes: When a black light is detected at an intersection, it is determined whether it is a true black light. If a colored traffic light is continuously detected within 5 seconds, it is determined to be a false black light. For truly dark lights, a pass-through decision is made; The decision logic for a false black light is to use the corresponding color signal light.

7. The decision-making and control method for traffic light intersections according to claim 1, characterized in that, The method further includes: When an intersection has both a straight-ahead round light and a left-turn arrow light, and the left-turn arrow light is black, the color of the straight-ahead round light is used to make the passage decision.

8. A decision-making and control device for a traffic light intersection, characterized in that, include: The first state transition unit is used to transition the decision state machine to the near-stop line state when the real-time distance between the vehicle and the stop line of the intersection ahead is less than or equal to a preset distance threshold. The green flashing passage decision unit is used to make a passage decision based on the currently perceived green flashing / yellow light status when approaching the stop line, and to determine the current state of the state machine based on the obtained passage decision and / or traffic light status. The current state includes: approaching the stop line, red light stop, and light obstruction. The second state transition unit is used to transition the state machine from the approach to the stop line state to the intersection passage state when the vehicle passes the stop line and enters the intersection ahead. The intersection passage decision unit is used to make low-speed intersection passage decisions based on the traffic light status, vehicle position and speed during the intersection passage state, until the vehicle leaves the intersection ahead.

9. An electronic device comprising a memory, a processor, and a computer program stored in the memory and running on the processor, characterized in that, When the processor executes the computer program, it implements the decision control method for traffic light intersections as described in any one of claims 1 to 7.

10. A non-transitory computer-readable storage medium having a computer program stored thereon, characterized in that, When the computer program is executed by the processor, it implements the decision control method for traffic light intersections as described in any one of claims 1 to 7.

Images