Traffic light passing decision-making method

By using part of Markov decision-making process and state machine in intelligent driving technology, combined with image recognition technology, the problem of intelligent driving technology's dependence on networked information is solved, the accuracy and stability of traffic light recognition are improved, and the accuracy and stability of traffic lights are improved. The safety and intelligence level of autonomous driving vehicles.

CN119068689APending Publication Date: 2024-12-03ANHUI YUCHI INTELLIGENT TECH CO LTD
View PDF 0 Cites 0 Cited by

Patent Information

Application Number
CN202411344434.3
Authority / Receiving Office
CN · China
Patent Type
Applications(China)
Current Assignee / Owner
Filing Date
2024-09-25
Publication Date
2024-12-03

AI Technical Summary

Technical Problem

The existing intelligent driving technology relies on high-cost networked information in traffic light recognition, resulting in insufficient identification accuracy and stability, especially in complex environments that are susceptible to external interference, affecting the traffic safety of autonomous driving vehicles.

Method used

The partially observable Markov decision-making process (POMDP) ​​is used to combine the state machine to reduce dependence on networked information through image recognition and multi-frame image processing, improve the recognition stability and accuracy of traffic lights, and reduce costs.

Benefits of technology

In complex application scenarios, traffic signal recognition capabilities and decision-making stability have been significantly improved, external interference has been reduced, and the intelligence and safety of autonomous driving technology have been improved, ensuring that vehicles can accurately identify and comply with traffic regulations, and reducing the risk of running red lights.

✦ Generated by Eureka AI based on patent content.

Smart Images

  • Figure CN119068689A_ABST
    Figure CN119068689A_ABST
Patent Text Reader

Abstract

The invention discloses a traffic light passing decision-making method. The traffic light passing decision-making method comprises the steps that the color state of a current traffic light is judged through a partially observable Markov decision, and the color state, judged through the partially observable Markov decision, of the current traffic light is called by a state machine; and determining the state of the state machine and determining a passing decision based on the state of the state machine. According to the traffic light passing decision method, a light-weight and high-robustness traffic light passing decision is provided, dependence on traffic light networking information can be reduced, and meanwhile, the traffic signal recognition capability and decision stability of an automatic driving vehicle in a complex application scene can be improved; therefore, the overall intelligent level and safety of the automatic driving technology are improved.
Need to check novelty before this filing date? Find Prior Art

Description

Technical Field

[0001] The present application relates to the field of intelligent driving, and in particular to a traffic light passage decision-making method. Background Art

[0002] Current intelligent driving technology usually relies on the interaction between vehicles and traffic infrastructure in terms of traffic light recognition to enhance the vehicle's perception and decision-making capabilities. The interaction information between vehicles and traffic infrastructure usually comes from networked information, such as traffic management departments, large map service providers, or direct access to the road network through sensors to receive traffic light information. Although the above-mentioned use of the interactive information between vehicles and traffic infrastructure from networked information to achieve traffic light recognition can improve the accuracy and real-time performance of traffic light recognition, it also leads to high costs, making it difficult to popularize such traffic light recognition technology and making it difficult for all car owners to enjoy the convenience brought by such traffic light recognition technology.

[0003] As a lower-cost alternative, the traffic light signal solution based on image recognition can avoid some infrastructure investment, but in actual application it is easily interfered by various external factors, which seriously limits the ability of autonomous driving vehicles to pass through intersections. For example, on rainy days, the monocular camera used by the current intelligent driving system is easily blocked by raindrops and wipers, resulting in some frames being unable to perceive stable traffic light information from the scene, which greatly reduces the stability and safety of the autonomous driving system at intersections, requiring the driver to take over frequently, which may lead to dangerous situations such as running red lights, making the autonomous driving technology appear to be not smart and reliable enough.

[0004] Based on the above problems, there is an urgent need for a solution that can not only reduce the dependence on traffic light network information, but also significantly improve the real-time and stability of traffic light recognition. Summary of the invention

[0005] One advantage of the present application is that it provides a traffic light passage decision method, wherein the traffic light passage decision method provides a lightweight and highly robust traffic light passage decision, which can reduce the dependence on traffic light network information while improving the traffic signal recognition ability and decision-making stability of autonomous driving vehicles in complex application scenarios, thereby improving the overall intelligence level and safety of autonomous driving technology.

[0006] One advantage of the present application is that it provides a traffic light passage decision method, wherein the traffic light passage decision method uses a partially observable Markov decision process (Partially Observable Markov Decision Process, abbreviated as: POMDP) ​​to realize the recognition of signal lights, and defines multiple vehicle position events and multiple traffic light events through a state machine. It can not only ensure the accuracy of passage decisions in simple application scenarios (for example, there is only one traffic light), but also ensure the accuracy of passage decisions in more complex application scenarios (for example, when the green light is about to turn red; at an intersection with multiple traffic lights; at an intersection with a waiting area).

[0007] One advantage of the present application is that it provides a traffic light passage decision method, wherein the traffic light passage decision method uses a partially observable Markov decision process (Partially Observable Markov Decision Process, abbreviated as: POMDP) ​​to realize the recognition of traffic lights, which can improve the stability of recognition and reduce external interference caused by obstructions such as rain, wipers, and fog.

[0008] An advantage of the present application is that it provides a traffic light passing decision method, wherein a part of the observable Markov decision process in the traffic light passing decision method is simplified and generalized, which can reduce the cost to a certain extent.

[0009] According to one aspect of the present application, a traffic light passing decision method is provided, comprising the steps of:

[0010] Determine the color state of the current traffic light by partially observable Markov decision making, wherein the color state of the current traffic light determined by the partially observable Markov decision making is called by the state machine;

[0011] Determining the state of the state machine and determining a traffic decision based on the state of the state machine includes the steps of:

[0012] Determine whether to activate the dual-light state machine. In response to the current state machine being in a non-idle state, submit the vehicle position event and the traffic light event to the state machine for processing; and determine the traffic light color and the parking position according to the state returned by the state machine; in response to the current state machine being in an idle state and reading the dual-light situation from the map information, activate the state machine and submit the vehicle position event and the traffic light event to the state machine for processing, and the first activation of the state machine regards the traffic light as a green light and does not set the parking position; in response to the current state machine being in an idle state and reading only a single light from the map information, obtain the color state of the current traffic light determined by the partially observable Markov decision, and determine whether to stop based on the color state of the current traffic light determined by the partially observable Markov decision.

[0013] In one implementation of the traffic light passage decision method according to an embodiment of the present application, the color state of the current traffic light is determined by partially observable Markov decision, including the steps of: obtaining a multi-frame image containing traffic lights, and identifying all traffic light information in the current frame image; defining two delay arrays and two index variables, and initializing the two delay arrays and the two index variables; obtaining the color state of the traffic lights and the number of traffic lights, and updating the counts of red lights and green lights based on the traffic light type; in response to the sum of the count of red lights and the count of green lights in the delay array being less than a preset count threshold, continuing to update the counts of red lights and green lights; in response to the sum of the count of red lights and the count of green lights in the delay array being greater than or equal to the preset count threshold, updating the count of red lights and the count of green lights according to the color state of the light in the current frame and the color state of the light in the previous frame; judging the color state of the current traffic light based on the final count of red lights and the final count of green lights; returning to the step of obtaining a multi-frame image containing traffic lights, and identifying the signal states of all traffic lights in the current frame image.

[0014] In one implementation of the traffic light passage decision method according to an embodiment of the present application, in response to the sum of the red light count and the green light count in the delay array being greater than or equal to a preset count threshold, the red light count and the green light count are updated according to the state of the light in the current frame and the state of the light in the previous frame, including the steps of: in response to the color state of the light in the previous frame being red and the color state of the light in the current frame being green, increasing the red light count; and in response to the color state of the light in the previous frame being green and the color state of the light in the current frame being red, increasing the green light count.

[0015] In one implementation of the traffic light passage decision method according to an embodiment of the present application, the color state of the current traffic light is judged based on the final red light count and the final green light count, including the steps of: in response to the final red light count being greater than or equal to the final green light count, determining that the color of the current traffic light is red; and in response to the final red light count being less than the final green light count, determining that the color of the current traffic light is green.

[0016] In one implementation of the traffic light passage decision method according to the embodiment of the present application, in response to the current state machine being in a non-idle state, the vehicle position event and the traffic light event are submitted to the state machine for processing; and the traffic light color and the parking position are determined according to the state returned by the state machine, including the steps of: in response to the state returned by the state machine indicating that it is necessary to stop at the stop line, the traffic light color is determined to be red, and the position in front of the sidewalk in the map is taken as the parking position; in response to the state returned by the state machine indicating that it is necessary to stop in the waiting area, the traffic light color is determined to be red, and the position in the waiting area in the map is taken as the parking position; in response to the state returned by the state machine being a state other than indicating the need to stop at the stop line and the need to stop in the waiting area, the traffic light is determined to be green, and no parking position is set.

[0017] In one implementation of the traffic light passage decision method according to an embodiment of the present application, the traffic light passage decision method also includes the steps of: creating a vehicle position event and creating a traffic light event, wherein three vehicle position events are created through a state machine according to the vehicle position, and the three vehicle position events are: the vehicle is in front of the waiting area, the vehicle is in the waiting area, and the vehicle has passed the waiting area; six traffic light events are created through a state machine according to the vehicle position and the color state of the current traffic light determined by a partially observable Markov decision, and the six traffic light events are: red light for left turn and red light for going straight, red light for left turn and green light for going straight, green light for left turn and red light for going straight, green light for left turn and green light for going straight, red light for left turn and green light for going straight, red light for left turn and green light for going left, red light for left turn, and green light for left turn.

[0018] In one implementation of the traffic light passage decision method according to an embodiment of the present application, in response to the vehicle position event that the vehicle is in front of the waiting area, event status information is initialized, wherein the initialized event status information is that the left turn light is green.

[0019] In one implementation of the traffic light passage decision method according to an embodiment of the present application, in response to the color state update of the current traffic light, the state of the vehicle is updated according to the traffic light event, wherein in response to the traffic light event that: the light is red for the left turn and the light is red for the straight line, it is confirmed that the vehicle needs to stop at the stop line; in response to the traffic light event that: the light is red for the left turn and the light is green for the straight line, it is confirmed that the vehicle needs to stop in the waiting area; in response to the traffic light event that: other traffic light events except the light is red for the left turn and the light is green for the straight line and the light is red for the left turn and the light is red for the straight line, it is determined that the vehicle can pass.

[0020] In one implementation of the traffic light passage decision method according to an embodiment of the present application, in response to a state machine indicating that a stop is required at a stop line and the vehicle position event is that the vehicle is before a waiting area, the color state of the current traffic light is updated; in response to a state machine indicating that a stop is required at a stop line and the vehicle position event is that the vehicle has passed through the waiting area, and the traffic light event is a green light for a left turn, it is determined that the vehicle is passable; in response to a state machine indicating that a stop is required at a stop line and the vehicle position event is that the vehicle has passed through the waiting area, and the traffic light event is a traffic light event other than a green light for a left turn, it is determined that the vehicle needs to stop in the waiting area; in response to a state machine indicating that a stop is required at a stop line and the vehicle position event is that the vehicle has passed through the waiting area, and the traffic light event is a traffic light event other than a green light for a left turn, it is determined that the vehicle needs to stop in the waiting area; in response to a state machine indicating that a stop is required at a stop line and the vehicle position event is that the vehicle is before the waiting area, it is determined that the vehicle is in a pending state.

[0021] In one implementation of the traffic light passage decision method according to an embodiment of the present application, in response to the state machine indicating that it is necessary to stop in the waiting area, and the vehicle position event is that the vehicle is in the waiting area or the vehicle has passed the waiting area, the color state of the current traffic light is updated; in response to the state machine indicating that it is necessary to stop in the waiting area, and the vehicle position event is that the vehicle is in front of the waiting area, the vehicle is determined to be in a pending state.

[0022] Further objectives and advantages of the present application will be fully reflected through understanding of the following description and drawings.

[0023] These and other objects, features and advantages of the present application are fully reflected in the following detailed description, drawings and claims. BRIEF DESCRIPTION OF THE DRAWINGS

[0024] By describing the embodiments of the present application in more detail in conjunction with the accompanying drawings, the above and other purposes, features and advantages of the present application will become more apparent. The accompanying drawings are used to provide a further understanding of the embodiments of the present application and constitute a part of the specification. Together with the embodiments of the present application, they are used to explain the present application and do not constitute a limitation of the present application. In the drawings, the same reference numerals generally represent the same components or.

[0025] Figure 1 A flowchart schematic diagram of a traffic light passing decision method according to an embodiment of the present application is illustrated.

[0026] Figure 2 The figure illustrates a flowchart schematic diagram of a process of determining the color state of the current traffic light through partially observable Markov decision making in a traffic light passage decision method according to an embodiment of the present application.

[0027] Figure 3 The figure illustrates a flowchart schematic diagram of a process of determining a traffic decision by a state machine in a traffic light traffic decision method according to an embodiment of the present application. DETAILED DESCRIPTION

[0028] Below, the exemplary embodiments according to the present application will be described in detail with reference to the accompanying drawings. Obviously, the described embodiments are only part of the embodiments of the present application, rather than all the embodiments of the present application, and it should be understood that the present application is not limited to the exemplary embodiments described here.

[0029] It is understood that the term "one" should be understood as "at least one" or "one or more", that is, in one embodiment, the number of an element can be one, while in another embodiment, the number of the element can be multiple, and the term "one" cannot be understood as a limitation on the number. "Multiple" means greater than or equal to two.

[0030] Although ordinals such as "first," "second," and the like will be used to describe various components, those components are not limited herein. The term is used only to distinguish one component from another. For example, a first component may be referred to as a second component, and likewise, a second component may be referred to as a first component without departing from the teachings of the present application. The term "and / or" as used herein includes any and all combinations of one or more associated listed items.

[0031] The terms used herein are only used for the purpose of describing various embodiments and are not intended to be limiting. As used herein, the singular form is intended to include the plural form, unless the context clearly indicates an exception. It will also be understood that the terms "including" and / or "having" when used in this specification specify the presence of the described features, numbers, operations, components, elements, or combinations thereof, without excluding the presence or addition of one or more other features, numbers, operations, components, elements, or combinations thereof.

[0032] Schematic traffic light decision-making method

[0033] like Figures 1 to 3 As shown, the traffic light passage decision method according to the embodiment of the present application is explained. The traffic light passage decision method provides a lightweight and highly robust traffic light passage decision, which uses a partially observable Markov decision process (POMDP) ​​to realize the recognition of signal lights, determines the passage decision through a state machine, and can reduce the dependence on traffic light network information.

[0034] Accordingly, if Figure 1As shown, the traffic light passage decision method according to the embodiment of the present application includes the steps of: S110, determining the color state of the current traffic light by partially observable Markov decision, wherein the color state of the current traffic light determined by the partially observable Markov decision is called by the state machine; S120, determining the state of the state machine and determining the passage decision based on the state of the state machine, including the steps of: determining whether to activate the double-light state machine, in response to the current state machine being in a non-idle state, submitting the vehicle position event and the traffic light event to the state machine for processing; and determining the traffic light color and the parking position according to the state returned by the state machine; in response to the current state machine being in an idle state and reading the double-light situation from the map information at the same time, activating the state machine, and submitting the vehicle position event and the traffic light event to the state machine for processing, and the first activation of the state machine regards the traffic light as a green light, and does not set the parking position; in response to the current state machine being in an idle state and reading only a single light situation from the map information at the same time, obtaining the color state of the current traffic light determined by the partially observable Markov decision, and determining whether to stop based on the color state of the current traffic light determined by the partially observable Markov decision.

[0035] In step S110 , the color state of the current traffic light is determined by partially observable Markov decision making. Specifically, first, a multi-frame image containing traffic lights is obtained, and all traffic light information in the current frame image is identified, wherein the traffic light information includes the color state and / or countdown of the traffic lights, and the number of traffic lights; then, two delay arrays and two index variables are defined, and the two delay arrays and the two index variables are initialized; then, the traffic light color state and the number of traffic lights are obtained, and the counts of red lights and green lights are updated based on the traffic light type; in response to the sum of the count of red lights and the count of green lights in the delay array being less than a preset count threshold, the counts of red lights and green lights continue to be updated; in response to the sum of the count of red lights and the count of green lights in the delay array being greater than or equal to the preset count threshold, the count of red lights and the count of green lights are updated according to the color state of the lights in the current frame and the color state of the lights in the previous frame; then, based on the final count of red lights and the final count of green lights, the color state of the current traffic light is determined; finally, the step of returning to obtaining a multi-frame image containing traffic lights and identifying the signal states of all traffic lights in the current frame image is returned, so as to facilitate subsequent decision-making processing by a state machine based on traffic regulations.

[0036] More specifically, in the process of defining two delay arrays and two index variables, and initializing the two delay arrays and the two index variables, different delay arrays can be selected for use according to the traffic light type (for example, left turn, straight go); the corresponding index variable can be selected according to the color state of the traffic light type; wherein the two index variables correspond to the two delay arrays.

[0037] In the process of obtaining the traffic light color status and the number of traffic lights, and updating the counts of red and green lights based on the traffic light type, if the traffic light information is empty, that is, the number of traffic lights is 0; set the current position in the delay array to the default value, for example, the green light; if the traffic light information is not empty, obtain the traffic light information and update the counts of red and green lights based on the traffic light type.

[0038] In a specific example of the present application, if the number of traffic lights is 0, the value of the index position (index) in the delay array (array) is set to 1, which is considered to be a green light.

[0039] like Figure 2 As shown, according to convention, yellow light is represented by 0, green light is represented by 1, and red light is represented by 2. Set the value of the index position in the delay array according to the traffic light color type:

[0040] type == 1 (green light): the value of the index position is set to 1 (passable).

[0041] type == 0 (red light) or type == 2 (yellow light): the value of the index position is set to 0 (not passable).

[0042] After the setting is completed, the value of the index position is increased by 1 for the next value update.

[0043] According to the latest value in the current delay array, update the red light count red_cnt and the green light count green_cnt, and accumulate the number of 0 and 1 in the delay array to maintain the stable state of the traffic light. The traffic light switching logic is:

[0044] If the red light count red_cnt + the green light count green_cnt is less than 10, the corresponding color counts are accumulated; when the red light count red_cnt + the green light count green_cnt reaches 10, the change of the traffic light is determined according to the delay state of the current and previous frames. If the previous frame is red light (0) and the current frame is green light (1), the red light count red_cnt increases and the green light count green_cnt decreases; if the previous frame is green light (1) and the current frame is red light (0), the red light count red_cnt decreases and the green light count green_cnt increases.

[0045] In other words, in response to the sum of the red light count and the green light count in the delay array being less than a preset count threshold, the red light count and the green light count continue to be updated; in response to the sum of the red light count and the green light count in the delay array being greater than or equal to the preset count threshold, the red light count and the green light count are updated according to the color state of the light in the current frame and the color state of the light in the previous frame.

[0046] In response to the color state of the light in the previous frame being red and the color state of the light in the current frame being green, the count of the red light is increased; in response to the color state of the light in the previous frame being green and the color state of the light in the current frame being red, the count of the green light is increased.

[0047] In the process of judging the color state of the current traffic light based on the final red light count and the final green light count, the color state of the current traffic light is determined based on the comparison of the red light count and the green light count: if the red light count is greater than or equal to the green light count, it is set to red light (0), otherwise it is set to green light (1).

[0048] In other words, in response to the final red light count being greater than or equal to the final green light count, the color of the current traffic light is determined to be red; in response to the final red light count being less than the final green light count, the color of the current traffic light is determined to be green.

[0049] The traffic light decision method uses a partially observable Markov decision process (POMDP) ​​to recognize the signal light, which can improve the stability of recognition and reduce external interference caused by rain, wipers, fog, etc. According to experiments, at a specific intersection, when the rain just blocks the front view image, the accuracy of the color state of the current traffic light output by the partially observable Markov decision process is 98%.

[0050] The partially observable Markov decision process in the traffic light passing decision method is simplified and generalized, which can reduce the cost to a certain extent.

[0051] In the embodiment of the present application, multiple vehicle position events and multiple traffic light events are defined by the state machine, which can not only ensure the accuracy of traffic decisions in simple application scenarios (for example, there is only one traffic light), but also ensure the accuracy of traffic decisions in more complex application scenarios, such as when the green light is about to turn red; at an intersection with multiple traffic lights; at an intersection with a waiting area. For more complex traffic light intersections, there is a "protected left turn" scenario, which requires processing of multiple decisions on whether to pass directly, or to stop in front of the sidewalk, or to stop in the waiting area, as well as "passing the sidewalk stop line when the left turn is green, and the vehicle is in the waiting area, the left turn light turns red, and the vehicle should pass normally." In this way, in extreme cases, such as when the green light is about to turn red, it can also ensure that the vehicle passes in accordance with traffic regulations and prohibits running red lights, which greatly improves the safety and reliability of autonomous driving vehicles in complex traffic environments. Accordingly, the traffic light passage decision method can reduce the dependence on traffic light network information while improving the traffic signal recognition ability and decision-making stability of autonomous driving vehicles in complex application scenarios, thereby improving the overall intelligence level and safety of autonomous driving technology.

[0052] Accordingly, before step S120, a vehicle position event and a traffic light event may be created, wherein three vehicle position events are created through a state machine according to the vehicle position, and the three vehicle position events are: the vehicle is in front of the waiting area, the vehicle is in the waiting area, and the vehicle has passed the waiting area; six traffic light events are created through a state machine according to the vehicle position and the color state of the current traffic light determined by the partially observable Markov decision, and the six traffic light events are: left turn is red and the straight line is red, left turn is red and the straight line is green, left turn is green and the straight line is red, left turn is green and the straight line is green, left turn is red, and left turn is green. Event state information may be generated, wherein the event state message is state information spliced ​​by the vehicle position event and the traffic light event. The passage decision may be expressed by the event state information.

[0053] In step S120, the state of the state machine is determined and the traffic decision is determined based on the state of the state machine. Specifically, it is determined whether to activate the dual-light state machine. In response to the current state machine being in a non-idle state, the vehicle position event and the traffic light event are submitted to the state machine for processing; and the traffic light color and parking position are determined according to the state returned by the state machine; in response to the current state machine being in an idle state and the dual-light situation is read from the map information, the state machine is activated, and the vehicle position event and the traffic light event are submitted to the state machine for processing, and the first activation of the state machine is regarded as the traffic light being green, and the parking position is not set; in response to the current state machine being in an idle state and the single-light situation is read from the map information, the color state of the current traffic light determined by the partially observable Markov decision is obtained, and whether to stop is determined based on the color state of the current traffic light determined by the partially observable Markov decision.

[0054] In response to the current state machine being in a non-idle state, the vehicle position event and the traffic light event are submitted to the state machine for processing; and in the process of determining the traffic light color and the parking position according to the state returned by the state machine, in response to the state returned by the state machine indicating that it is necessary to stop at the stop line, the traffic light color is determined to be red, and the position in front of the sidewalk in the map is taken as the parking position; in response to the state returned by the state machine indicating that it is necessary to stop in the waiting area, the traffic light color is determined to be red, and the position in the waiting area in the map is taken as the parking position; in response to the state returned by the state machine being a state other than indicating the need to stop at the stop line and the need to stop in the waiting area, the traffic light is determined to be green, and no parking position is set.

[0055] The state machine generates travel decisions in the form of event status messages and prints them in the log to facilitate debugging.

[0056] After generating a passage decision, the vehicle behavior under different traffic light states processed by the state machine is returned, including the parking position and traffic light color.

[0057] The following is an introduction to the internal design logic of the state machine. Figure 3 In response to the vehicle position event that the vehicle is in front of the waiting area, the event status information is initialized, wherein the initialized event status information is that the left turn light is green.

[0058] In response to the current traffic light color status update, the vehicle status is updated according to the traffic light event, wherein in response to the traffic light event: the light is red for the left turn and the light is red for the straight line, it is confirmed that the vehicle needs to stop at the stop line; in response to the traffic light event: the light is red for the left turn and the light is green for the straight line, it is confirmed that the vehicle needs to stop in the waiting area; in response to the traffic light event: other traffic light events except the light is red for the left turn and the light is green for the straight line and the light is red for the left turn and the light is red for the straight line, it is determined that the vehicle can pass.

[0059] In response to the state machine indicating that it is necessary to stop at the stop line, and the vehicle position event is that the vehicle is before the waiting area, the color state of the current traffic light is updated; in response to the state machine indicating that it is necessary to stop at the stop line, and the vehicle position event is that the vehicle has passed the waiting area, and the traffic light event is a green light for a left turn, it is determined that the vehicle can pass; in response to the state machine indicating that it is necessary to stop at the stop line, and the vehicle position event is that the vehicle has passed the waiting area, and the traffic light event is a traffic light event other than a green light for a left turn, it is determined that the vehicle needs to stop in the waiting area; in response to the state machine indicating that it is necessary to stop at the stop line, and the vehicle position event is that the vehicle has passed the waiting area, and the traffic light event is a traffic light event other than a green light for a left turn, it is determined that the vehicle needs to stop in the waiting area; in response to the state machine indicating that it is necessary to stop at the stop line, and the vehicle position event is that the vehicle is before the waiting area, it is determined that the vehicle is in a pending state.

[0060] In response to the state machine indicating that it is necessary to stop in the waiting area, and the vehicle position event is that the vehicle is in the waiting area or the vehicle has passed the waiting area, the color state of the current traffic light is updated; in response to the state machine indicating that it is necessary to stop in the waiting area, and the vehicle position event is that the vehicle is before the waiting area, the vehicle is determined to be in a pending state.

[0061] In response to the state machine indicating that the vehicle can pass, and the vehicle position event is that the vehicle is in the waiting area or the vehicle has passed the waiting area, the vehicle state is maintained as a passable state; in response to the state machine indicating that the vehicle can pass, and the vehicle position event is that the vehicle is before the waiting area, the vehicle is determined to be in a pending state.

[0062] In summary, the traffic light passage decision method according to the embodiment of the present application is explained, and the traffic light passage decision method provides a lightweight and highly robust traffic light passage decision, which can reduce the dependence on traffic light network information while improving the traffic signal recognition ability and decision-making stability of the autonomous driving vehicle in complex application scenarios, thereby improving the overall intelligence level and safety of autonomous driving technology.

[0063] The present application and its implementation methods are described above, and such description is not restrictive. The drawings show only one implementation method of the present application, and the actual structure is not limited thereto. In short, if ordinary technicians in the field are inspired by it and design structural methods and embodiments similar to the technical solution without creative design without departing from the inventive purpose of the present application, they should all fall within the protection scope of the present application.

Claims

1. A traffic light decision-making method, It is characterized in that Includes steps: Determine the color state of the current traffic light by partially observable Markov decision making, wherein the color state of the current traffic light determined by the partially observable Markov decision making is called by the state machine; Determining the state of the state machine and determining a traffic decision based on the state of the state machine includes the steps of: Determine whether to activate the dual-light state machine. In response to the current state machine being in a non-idle state, submit the vehicle position event and the traffic light event to the state machine for processing; and determine the traffic light color and the parking position according to the state returned by the state machine; in response to the current state machine being in an idle state and reading the dual-light situation from the map information, activate the state machine and submit the vehicle position event and the traffic light event to the state machine for processing, and the first activation of the state machine regards the traffic light as a green light and does not set the parking position; in response to the current state machine being in an idle state and reading only a single light from the map information, obtain the color state of the current traffic light determined by the partially observable Markov decision, and determine whether to stop based on the color state of the current traffic light determined by the partially observable Markov decision.

2. The traffic light decision-making method according to claim 1, It is characterized in that The color state of the current traffic light is determined by partially observable Markov decision making, including the following steps: Obtain multiple frames of images containing traffic lights, and identify all traffic light information in the current frame image; Define two delay arrays and two index variables, and initialize the two delay arrays and two index variables; Obtaining the color state of the traffic lights and the number of traffic lights, and updating the counts of red lights and green lights based on the type of traffic lights; in response to the sum of the counts of red lights and green lights in the delay array being less than a preset count threshold, continuing to update the counts of red lights and green lights; in response to the sum of the counts of red lights and green lights in the delay array being greater than or equal to the preset count threshold, updating the counts of red lights and green lights according to the color state of the lights in the current frame and the color state of the lights in the previous frame; Based on the final red light count and the final green light count, determine the current color state of the traffic light; Return to the step of acquiring multiple frames of images containing traffic lights and identifying the signal states of all traffic lights in the current frame image.

3. The traffic light decision-making method according to claim 2, It is characterized in that In response to the sum of the count of red lights and the count of green lights in the delay array being greater than or equal to a preset count threshold, the count of red lights and the count of green lights are updated according to the state of lights in the current frame and the state of lights in the previous frame, comprising the steps of: In response to the color state of the light in the previous frame being a red light and the color state of the light in the current frame being a green light, increasing the count of the red light; and In response to the color state of the light in the previous frame being the green light and the color state of the light in the current frame being the red light, the count of the green light is increased.

4. The traffic light decision-making method according to claim 3, It is characterized in that Based on the final red light count and the final green light count, the color state of the current traffic light is determined, including the steps of: In response to the final red light count being greater than or equal to the final green light count, determining that the color of the current traffic light is red; and In response to the final red light count being less than the final green light count, it is determined that the color of the current traffic light is green.

5. The traffic light decision-making method according to claim 1, It is characterized in that In response to the current state machine being in a non-idle state, submitting a vehicle position event and a traffic light event to the state machine for processing; and determining a traffic light color and a parking position according to a state returned by the state machine, including the steps of: In response to the state returned by the state machine indicating that it is necessary to stop at the stop line, the traffic light color is determined to be red, and the position in front of the sidewalk in the map is taken as the parking position; In response to the state returned by the state machine indicating that it is necessary to stop in the waiting area, the traffic light color is determined to be red, and the position of the waiting area in the map is taken as the parking position; In response to the state returned by the state machine being a state other than indicating the need to stop at the stop line and the need to stop in the waiting area, it is determined that the traffic light is green and no parking position is set.

6. The traffic light decision-making method according to claim 4, It is characterized in that The traffic light passage decision method also includes the steps of: creating a vehicle position event and creating a traffic light event, wherein three vehicle position events are created through a state machine according to the vehicle position, and the three vehicle position events are: the vehicle is in front of a waiting area, the vehicle is in the waiting area, and the vehicle has passed the waiting area; six traffic light events are created through a state machine according to the vehicle position and the color state of the current traffic light determined by a partially observable Markov decision, and the six traffic light events are: the left turn is red and the straight line is red, the left turn is red and the straight line is green, the left turn is green and the straight line is red, the left turn is green and the straight line is green, the left turn is red, and the straight line is green, the left turn is red, and the left turn is green.

7. The traffic light decision-making method according to claim 6, It is characterized in that In response to the vehicle position event that the vehicle is in front of the waiting area for turning, event status information is initialized, wherein the initialized event status information is that the left turn light is green.

8. The traffic light passing decision method according to claim 6, It is characterized in that In response to the current traffic light color status update, the vehicle status is updated according to the traffic light event, wherein in response to the traffic light event: the light is red for the left turn and the light is red for the straight line, it is confirmed that the vehicle needs to stop at the stop line; in response to the traffic light event: the light is red for the left turn and the light is green for the straight line, it is confirmed that the vehicle needs to stop in the waiting area; in response to the traffic light event: other traffic light events except the light is red for the left turn and the light is green for the straight line and the light is red for the left turn and the light is red for the straight line, it is determined that the vehicle can pass.

9. The traffic light passing decision method according to claim 6, It is characterized in that In response to the state machine indicating that it is necessary to stop at the stop line, and the vehicle position event is that the vehicle is before the waiting area, the color state of the current traffic light is updated; in response to the state machine indicating that it is necessary to stop at the stop line, and the vehicle position event is that the vehicle has passed the waiting area, and the traffic light event is a green light for a left turn, it is determined that the vehicle can pass; in response to the state machine indicating that it is necessary to stop at the stop line, and the vehicle position event is that the vehicle has passed the waiting area, and the traffic light event is a traffic light event other than a green light for a left turn, it is determined that the vehicle needs to stop in the waiting area; in response to the state machine indicating that it is necessary to stop at the stop line, and the vehicle position event is that the vehicle has passed the waiting area, and the traffic light event is a traffic light event other than a green light for a left turn, it is determined that the vehicle needs to stop in the waiting area; in response to the state machine indicating that it is necessary to stop at the stop line, and the vehicle position event is that the vehicle is before the waiting area, it is determined that the vehicle is in a pending state.

10. The traffic light passing decision method according to claim 6, It is characterized in that In response to the state machine indicating that it is necessary to stop in the waiting area, and the vehicle position event is that the vehicle is in the waiting area or the vehicle has passed the waiting area, the color state of the current traffic light is updated; in response to the state machine indicating that it is necessary to stop in the waiting area, and the vehicle position event is that the vehicle is before the waiting area, the vehicle is determined to be in a pending state.