Countdown probability-based driving adjustment method and system for autonomous vehicle

Abstract

The invention discloses a countdown probability-based driving adjustment method and system for an automatic driving vehicle, and the method comprises the following steps: constructing a traffic light periodic table based on historical periodic data, recognizing the color and countdown information of a current traffic light, and if the historical periodic data of the current traffic light is recorded in the traffic light periodic table, judging whether the traffic light is a traffic light color or not; if yes, calculating the remaining time of the current traffic light color according to the historical cycle data, otherwise, collecting the cycle data of the current traffic light, and updating the traffic light periodic table; if the countdown information of the current traffic light is not recognized, predicting the switching probability of the light color of the current traffic light by using historical cycle data of other traffic lights in the road section where the current traffic light is located; and dynamically generating an optimal passing strategy according to the current traffic light color, the remaining time, the light color switching probability, the current vehicle speed, the road speed limit and the front vehicle state.

Description

Technical Field

[0001] The present invention relates to the field of autonomous driving control, and particularly to a method and system for adjusting the driving of an autonomous vehicle based on countdown probability. Background Art

[0002] With the rapid development of intelligent transportation systems and autonomous driving technologies, how to ensure the safe and efficient driving of autonomous vehicles in complex traffic environments has become one of the research hotspots. In actual road traffic, whether it is manual driving or autonomous driving, when a vehicle passes through an intersection controlled by traffic lights, it must make a traffic decision based on the state of the traffic lights.

[0003] In an autonomous driving system, the perception of traffic light states usually relies on computer vision technology. Images are collected through a camera to identify the color and state of traffic lights, and then a vehicle driving strategy is formulated. In recent years, with the emergence of vehicle-to-everything (V2X) technology, by communicating between vehicles and infrastructure, the state information of traffic lights can be directly obtained, providing another way for autonomous vehicles to obtain traffic light states.

[0004] However, at present, most traffic lights at intersections have not been connected to V2X devices, resulting in limited coverage and popularity of V2X technology in practical applications. Therefore, in most cases, autonomous vehicles still need to rely on the vehicle-mounted perception system to obtain traffic light state information. However, when the traffic lights are partially blocked by other vehicles, pedestrians, or environmental obstacles, resulting in the lack of countdown information of the traffic lights, it is difficult for the system to accurately predict the current state and remaining time of the traffic lights, affecting the decision-making efficiency and safety of the autonomous driving system. Summary of the Invention

[0005] The present invention proposes a method and system for adjusting the driving of an autonomous vehicle based on countdown probability, which solves the problem that it is difficult for existing autonomous driving systems to make traffic decisions when traffic lights are partially blocked.

[0006] To solve the above technical problems, the present invention provides a method for adjusting the driving of an autonomous vehicle based on countdown probability, including the following steps:

[0007] Step S1: Construct a traffic light cycle table based on historical cycle data, identify the light color and countdown information of the current traffic light. If the historical cycle data of the current traffic light is recorded in the traffic light cycle table, calculate the remaining time of the light color of the current traffic light according to the historical cycle data; otherwise, execute Step S2;

[0008] Step S2: Collect the cycle data of the current traffic light, update the traffic light cycle table. If the countdown information of the current traffic light is not identified in Step S1, execute Step S3; otherwise, execute Step S4;

[0009] Step S3: Predict the switching probability of the current traffic light color using the historical cycle data of other traffic lights in the section where the current traffic light is located;

[0010] Step S4: Dynamically generate an optimal passing strategy based on the current traffic light color, remaining time, probability of traffic light color switching, current vehicle speed, road speed limit, and the state of the vehicle in front.

[0011] Preferably, in step S1, a Universally Unique Identifier (UUID) is assigned to each traffic light in the vehicle map according to the geographical location of the traffic light, and a traffic light cycle table is constructed based on the UUID, geographical location, duration of each traffic light color, start time of each traffic light color, and the timestamp record of traffic light color switching.

[0012] Preferably, in step S1, the traffic light color and countdown information of the current traffic light are identified by a multi - focal camera, including the following steps: Determine the distance between the vehicle and the traffic light. When the distance between the vehicle and the traffic light is greater than a set threshold, the traffic light color and countdown information of the current traffic light are identified by a long - focal camera; otherwise, the traffic light color and countdown information of the current traffic light are identified by a short - focal camera.

[0013] Preferably, in step S1, calculating the remaining time of the current traffic light color according to the historical cycle data includes the following steps:

[0014] Step S11: Obtain the duration T of the yellow light, the start time t of the yellow light, the duration T of the red light, the start time t of the red light, the duration T of the green light, and the start time t of the green light in the historical cycle data from the traffic light cycle table according to the UUID of the current traffic light, and calculate the complete cycle T of the traffic light = T + T + T. If the current traffic light color is green, execute step S12; if it is red, execute step S13; otherwise, execute step S14; y of the yellow light, the start time t y of the yellow light, the duration T r of the red light, the start time t r of the red light, the duration T g of the green light, and the start time t g of the green light, calculate the complete cycle T of the traffic light = T y + T r + T g , if the current traffic light color is green, then execute step S12; if it is red, then execute step S13; otherwise, execute step S14;

[0015] Step S12: Calculate the difference Δt = t - t between the current timestamp and the start time of the green light, calculate the time t when the current traffic light is in the starting state with the green light as the cycle start = Δt % T; Determine whether t g is less than the duration T s of the green light. If t s is less than T g , if t s < T g , then the current traffic light color is green, and the remaining time is t = Tg -t s ; If t s ≥T g , then determine if t s is less than T g +T y . If t s <T g +T y , then the light color of the current traffic light is yellow, and the remaining time is t = T g +T y -t s ; If t s ≥T g +T y , then the light color of the current traffic light is red, and the remaining time is t = T - t s ;

[0016] Step S13: Calculate the difference Δt' = t - t between the current timestamp and the start time of the red light r , calculate the time t s ' = Δt' % T when the current traffic light is in the starting state of the red light cycle; Determine if t s ' is less than the red light cycle T r . If t s '<T r , then the light color of the current traffic light is red, and the remaining time is t' = T r -t s '; If t s '≥T r , then determine if t s ' is less than T g +T r . If t s '<T g +T r , then the light color of the current traffic light is green, and the remaining time is t' = T g +T r -t s ; If t s ≥T g +T r , then the light color of the current traffic light is yellow, and the remaining time is t' = T - t s ';

[0017] Step S14: Calculate the difference Δt” = t - t between the current timestamp and the start time of the yellow light y , calculate the time t s ” = Δt” % T when the current traffic light is in the starting state of the yellow light cycle; Determine if t s ” is less than the yellow light cycle T y . If t s ”<Ty If the current traffic light color is yellow and the remaining time is t” = T y -t s ”; If t s ”≥T y , then determine if t s ” is less than T y +T r . If t s ”<T y +T r , then the current traffic light color is red and the remaining time is t” = T y +T r -t s ; If t s ≥T y +T r , then the current traffic light color is green and the remaining time is t” = T - t s ”.

[0018] Preferably, the step of collecting the cycle data of the current traffic light and updating the traffic light cycle table in step S2 includes the following steps:

[0019] Step S21: Obtain the UUID of the current traffic light and determine the traffic light color switching state. If the current traffic light changes from green to yellow, execute step S22; if the current traffic light changes from yellow to red, execute step S23; if the current traffic light changes from red to green, execute step S24;

[0020] Step S22: Record the time stamps of the start time of the yellow light, the time when the yellow light changes to red, and the time when the red light changes to green, calculate the cycles of the yellow light and the red light, and execute step S25;

[0021] Step S23: Record the time stamps of the start time of the red light and the time when the red light changes to green, calculate the cycle of the red light, and execute step S25;

[0022] Step S24: Record the time stamp of the start time of the green light;

[0023] Step S25: Determine whether the vehicle is passing the current traffic light for the second time according to the UUID of the current traffic light. If so, calculate the difference between the recorded start time stamps of the two lights of the same color, divide the difference of the time stamps by n, and determine whether the remainder is close to 0 and whether the quotient is greater than the sum of the cycles of the yellow light and the red light. If so, update the traffic light cycle table; otherwise, set n = n + 1;

[0024] Step S26: Repeat step S25 until the remainder of the difference between the time stamps of the start times of all traffic light colors divided by n is close to 0 and the quotient is greater than the sum of the cycles of the yellow light and the red light;

[0025] Step S27: Update the traffic light cycle table according to the UUID, geographical location, duration of each light color, start time of each light color, and timestamp of light color switching of the current traffic light.

[0026] Preferably, step S3 includes the following steps:

[0027] Step S31: Obtain the average value and standard deviation of the duration of each light color of other traffic lights on the section where the current traffic light is located;

[0028] Step S32: Assume that the duration of each light color of the traffic light follows a normal distribution, and calculate the likelihood function P(t s | status unchanged):

[0029]

[0030] In the formula, t s is the duration of the light color; σ s is the standard deviation of the duration of the light color; T s is the average value of the duration of the light color;

[0031] Step S33: Calculate the probability P(status about to change|t s ) that the light color of the current traffic light is about to change when the light color s of the current traffic light has lasted for t s :

[0032]

[0033] P(t s | status about to change) = 1 - P(t s | status unchanged);

[0034]

[0035] P(status unchanged) = 1 - P(status about to change);

[0036] P(t s ) = P(t s | status unchanged)P(status unchanged) + P(t s | status about to change)P(status about to change);

[0037] In the above formula, P(t s | status about to change) is the probability of observing a countdown of t s under the condition that the status is about to change; P(status about to change) is the probability that the status is about to change unconditionally; P(t s | status unchanged) is the countdown of t sThe probability; P(Status unchanged) is the probability that the status remains the current status without change under unconditional circumstances.

[0038] Preferably, step S4 includes the following steps:

[0039] Step S41: Determine whether the light color of the current traffic light is green. If it is, execute step S42; otherwise, execute step S49.

[0040] Step S42: Determine whether countdown information is recognized. If it is, execute step S43; otherwise, execute step S46.

[0041] Step S43: Calculate the minimum passing speed v of the vehicle min :

[0042]

[0043] In the formula, d is the distance of the vehicle from the stop line of the intersection; t is the remaining time of the green light.

[0044] Step S44: Determine whether the minimum passing speed is less than the road speed limit v max , if v min <v max

[0045] Then execute step S45; otherwise, execute step S49.

[0046] Step S45: If there is a vehicle in front of the vehicle and the speed of the vehicle in front is greater than the minimum passing speed v min , then the vehicle passes at a speed of v min <v<v max , otherwise execute step S49.

[0047] Step S46: If the probability that the predicted light color is about to turn green is less than the set probability threshold, execute step S49; otherwise, execute step S47.

[0048] Step S47: If there is a vehicle in front of the vehicle and the speed of the vehicle in front is less than the road speed limit, the vehicle passes at the speed of the vehicle in front; otherwise, it passes at the road speed limit.

[0049] Step S48: Determine whether the vehicle has passed the intersection. If it has, end the judgment; otherwise, return to step S41 for re-judgment.

[0050] Step S49: Decelerate and stop, return to step S41, and wait for the light color of the current traffic light to turn green.

[0051] The present invention also provides a driving adjustment system for an autonomous vehicle based on countdown probability, which is implemented based on the above-mentioned driving adjustment method for an autonomous vehicle based on countdown probability, and includes: a traffic light detection and status recognition module, a traffic light cycle table construction and update module, a Bayesian probability calculation module, a passing strategy calculation module, and a vehicle control module;

[0052] The traffic light detection and status recognition module: detects the traffic light status in real time through a multi-focal length camera, and recognizes the countdown information of the traffic light through a target detection model and OCR technology;

[0053] The traffic light cycle table construction and update module: constructs a traffic light cycle table by recording the timestamps of traffic light status changes, the duration of each light color, and the start timestamp multiple times, and dynamically updates the cycle table according to the real-time detection results;

[0054] The Bayesian probability calculation module: assumes that the duration of the light color follows a normal distribution, and calculates the probability of the traffic light status remaining or changing based on the historical data in the traffic light cycle table and the time that the current light color has lasted;

[0055] The passing strategy calculation module: dynamically generates an optimal passing strategy according to the current traffic light color, countdown information, vehicle's current position, speed, speed limit, and the status of the vehicle in front;

[0056] The vehicle control module: adjusts the vehicle's speed in real time according to the output of the passing strategy calculation module.

[0057] Preferably, the system further includes a data storage and feedback module, which is used to store the detection results of traffic lights, traffic light cycle table data, Bayesian probability calculation results, and passing strategy execution situations, dynamically update the traffic light cycle table data, and optimize the passing strategy.

[0058] Preferably, the system further includes a human-machine interaction module, which displays the current traffic light status, countdown information, passing strategy, and system operation status to the driver, and provides voice or visual prompts for the driver.

[0059] The advantages of the present invention at least include:

[0060] 1. By constructing a traffic light cycle table and using historical cycle data to calculate the remaining time of the current light color, it is possible to make full use of the accumulated information in the past and provide a relatively accurate time basis for the passing strategy; when encountering new traffic light cycle data, it can collect and update the traffic light cycle table in time, enabling the system to have the ability to adapt to changes in traffic conditions;

[0061] 2. In the case where the countdown information is not recognized, the historical cycle data of other surrounding traffic lights are used to predict the switching probability of the current traffic light color. Even when the traffic light countdown display is damaged, it can still provide decision-making reference for drivers;

[0062] 3. Dynamically generate an optimal traffic strategy based on multiple factors, which helps vehicles reasonably plan the driving speed and mode, reduce unnecessary stops and starts, and improve the overall road traffic efficiency. BRIEF DESCRIPTION OF THE DRAWINGS

[0063] Figure 1 It is a schematic flowchart of the method according to an embodiment of the present invention;

[0064] Figure 2 It is a schematic flowchart of the traffic light cycle analysis method according to an embodiment of the present invention;

[0065] Figure 3 It is a schematic flowchart of the traffic light color derivation process according to an embodiment of the present invention;

[0066] Figure 4 It is a schematic flowchart of the traffic light color prediction process according to an embodiment of the present invention;

[0067] Figure 5 It is a schematic flowchart of the vehicle traffic strategy process according to an embodiment of the present invention;

[0068] Figure 6 It is a schematic diagram of the hardware structure according to an embodiment of the present invention.

[0069] In the figure: 1 - long - focal camera; 2 - short - focal camera; 3 - camera signal line; 4 - domain controller; 5 - vehicle. DETAILED DESCRIPTION OF THE EMBODIMENTS

[0070] Next, the technical solutions in the embodiments of the present invention will be clearly and completely described in conjunction with the accompanying drawings in the embodiments of the present invention. Obviously, the described embodiments are only a part of the embodiments of the present invention, rather than all the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those of ordinary skill in the art without creative efforts belong to the protection scope of the present invention.

[0071] Accurately obtaining the future state changes of traffic lights in advance is of great significance for autonomous vehicles to formulate reasonable traffic strategies. By predicting the changes of traffic lights, vehicles can better plan the driving path and speed, improving the safety and comfort of driving. For example, if it can be known in advance that the traffic light is about to change from green to red, the vehicle can appropriately decelerate to avoid the discomfort caused by sudden braking; conversely, if the traffic light is about to turn green, the vehicle can accelerate in advance to reduce the waiting time and improve the driving efficiency.

[0072] During the current traffic light passing decision-making process of autonomous vehicles, challenges such as insufficient perception accuracy and limited information acquisition channels are faced. There is an urgent need to develop more advanced technical means to improve the reliability of traffic light status detection and achieve effective prediction of traffic light status changes, so as to optimize the driving strategies of autonomous vehicles and ensure their safe and efficient driving in complex traffic environments.

[0073] Based on the above background, embodiments of the present invention provide a Figure 1 driving adjustment method for autonomous vehicles based on countdown probability as shown in

[0074] Step S1: Construct a traffic light cycle table based on historical cycle data, identify the light color and countdown information of the current traffic light. If the historical cycle data of the current traffic light is recorded in the traffic light cycle table, calculate the remaining time of the light color of the current traffic light according to the historical cycle data; otherwise, execute Step S2.

[0075] Step S2: Collect the cycle data of the current traffic light and update the traffic light cycle table. If the countdown information of the current traffic light is not identified in Step S1, execute Step S3; otherwise, execute Step S4.

[0076] Step S3: Use the historical cycle data of other traffic lights in the section where the current traffic light is located to predict the switching probability of the light color of the current traffic light.

[0077] Specifically, through the vehicle's GPS positioning information and the data of the navigation map, the distance of the vehicle from the intersection is judged in real time, and a unique ID identifier UUID (Universally Unique Identifier) is assigned to the traffic light at this intersection.

[0078] The UUID of the traffic light can be generated in the following two modes:

[0079] (1) In areas with navigation map data, through the prior information of the navigation map, a fixed UUID is pre-generated. For example, hash the combined information "intersection ID + traffic light position + coordinates after precision processing" to generate a unique UUID for each traffic light;

[0080] (2) In areas without navigation map data, through the vehicle's GPS positioning information, combine the sensed coordinates to obtain the coordinates of the corresponding traffic light in the global coordinates, and hash the combined information "intersection ID + traffic light position + coordinates after precision processing" to generate a unique UUID for each traffic light.

[0081] UUID matching method:

[0082] (1) Areas with navigation map data: In the map data, a unique UUID is pre-assigned to each traffic light. After the vehicle system matches the GPS with the map, it directly reads the existing UUID, which can avoid potential conflicts when generating UUIDs by itself.

[0083] (2) Scenarios without navigation maps:

[0084] Error correction and fault tolerance: Considering the positioning error of GPS and the position recognition error of perception, a clustering algorithm is used to regard the traffic lights detected within a certain radius as the same entity, avoiding the wrong UUID matching due to errors.

[0085] Real-time verification: By comparing the physical characteristics of the traffic lights obtained by the camera in consecutive frames with the GPS positioning of the vehicle or the historical map data for secondary confirmation, and then matching the UUID.

[0086] During the vehicle's driving process, the system dynamically selects an appropriate camera focal length. When the vehicle is far from the intersection, the system enables a long-focus camera. The long-focus lens can effectively magnify distant objects, enabling clear images of traffic lights to be obtained even when the vehicle is far from the intersection. When the vehicle approaches the intersection, the system determines that it has entered the area near the intersection and uses a short-focus camera for image acquisition. Since the short-focus lens can capture clearer images at close range, this makes the traffic light box occupy more pixels in the image, thereby improving the detection accuracy and recognition accuracy of the traffic light status.

[0087] The switching of cameras needs to be comprehensively evaluated and judged based on factors such as the installation position of the in-vehicle camera, the pixel of the camera, and the focal length of the lens: First, continuously collect images of the same traffic light from far to near taken by two cameras at the same time, record the corresponding distance of each image from the camera, and obtain the distance range of the corresponding traffic light that each camera can observe. Use an object detection algorithm such as YOLO to detect the color of the traffic light and record the distance range that each camera can correctly and stably detect. If the detection ranges of the two cameras have an intersection, take the middle value of the intersection as the switching threshold; otherwise, adjust the camera model or installation position.

[0088] This scheme of switching multi-focal length cameras based on distance can make the traffic lights in the image occupy more effective pixels, reduce the recognition difficulty caused by insufficient resolution, especially in complex environments or bad weather conditions, and can significantly improve the accuracy and stability of image recognition, thus providing a reliable basis for subsequent vehicle passing decisions.

[0089] Use the YOLO algorithm or other object detection models to detect the light box and light color of the current traffic light. Retrieve according to the UUID of the current traffic light in the traffic light cycle table. If the cycle data of the current traffic light exists in the traffic light cycle table, deduce the remaining time of the current traffic light based on the cycle data.

[0090] When deducing the remaining time of the traffic light, it is necessary to further determine whether the countdown information of the current traffic light is recognized. If the countdown information is recognized, use the OCR algorithm to obtain the countdown of the current traffic light, record the current UNIX time stamp (UNIX Time Stamp) and the UUID, light color, countdown time, etc. of the current traffic light. Otherwise, perform cycle analysis and update the traffic light cycle table.

[0091] As Figure 2 shown, the cycle analysis of the traffic light includes the following steps:

[0092] Step 1: Obtain the UUID of the current traffic light, judge the light color of the current traffic light. If the current traffic light changes from green to yellow, execute Step 2. If the current traffic light changes from yellow to red, execute Step 3. If the current traffic light changes from red to green, execute Step 4;

[0093] Step 2: Record the time stamp of the start time of the yellow light The time stamp when the yellow light changes to red and the time stamp when the red light changes to green Calculate the cycle T of the yellow light y and the cycle T of the red light r :

[0094]

[0095] Proceed to Step 5;

[0096] Step 3: Record the time stamp of the start time of the red light The time stamp when the red light changes to green Calculate the cycle T of the red light r :

[0097]

[0098] Proceed to Step 5;

[0099] Step 4: Record the time stamp of the start time of the green light

[0100] Step 5: Determine whether the vehicle is passing the current traffic light for the second time based on the UUID of the current traffic light. If so, calculate the difference between the starting timestamps of the two same-color lights recorded, divide the difference in timestamps by n, and determine whether the remainder is close to 0 and whether the quotient is greater than the sum of the yellow light period and the red light period. If so, update the traffic light period table; otherwise, set n = n + 1;

[0101] Step 6: Repeat Step 5 until the remainder of the difference in timestamps of the starting times of all light colors divided by n is close to 0 and the quotient is greater than the sum of the yellow light period and the red light period. Then it can be determined that the same-color lights of the traffic lights obtained twice can be divided by the smallest integer and the error is acceptable, and at the same time, the limit of the minimum traffic light period is also satisfied;

[0102] Step 7: Use the quotient q as the period T of the current traffic light; use the T obtained in Step 2 y as the yellow light period, or set the yellow light period to the conventional 3 seconds; use the T calculated in Step 2 or Step 3 r as the red light period; set T g = T - T y - T r as the green light period; at the same time, record the starting timestamps of each light and update the traffic light period table.

[0103] After completing the period analysis of the traffic light, deduce the remaining time of the current traffic light. As shown in Figure 3 , it includes the following steps:

[0104] Step a: Obtain the duration T of the yellow light y , the starting time t of the yellow light y , the duration T of the red light r , the starting time t of the red light r , the duration T of the green light g and the starting time t of the green light g from the traffic light period table according to the UUID of the current traffic light, and calculate the complete period T of the traffic light = T y + T r + T g . If the light color of the current traffic light is green, execute Step b; if it is red, execute Step c; otherwise, execute Step d.

[0105] Step b: Calculate the difference Δt = t - t between the current timestamp and the starting time of the green light g , and calculate the time t when the current traffic light is in the starting state with the green light as the period s = Δt % T.

[0106] Judge whether t s is less than the duration T of the green light g . If t s < Tg , the current traffic light color is green, and the remaining time is t = T g -t s ; If t s ≥T g , then judge whether t s is less than T g +T y , if t s <T g +T y , the current traffic light color is yellow, and the remaining time is t = T g +T y -t s ; If t s ≥T g +T y , the current traffic light color is red, and the remaining time is t = T - t s .

[0107] Step c: Calculate the difference Δt' = t - t between the current timestamp and the start time of the red light r , calculate the time t s ' = Δt' % T when the current traffic light is in the starting state of the red light cycle.

[0108] Judge whether t s ' is less than the red light cycle T r , if t s '<T r , the current traffic light color is red, and the remaining time is t' = T r -t s '; If t s '≥T r , then judge whether t s ' is less than T g +T r , if t s '<T g +T r , the current traffic light color is green, and the remaining time is t' = T g +T r -t s ; If t s ≥T g +T r , the current traffic light color is yellow, and the remaining time is t' = T - t s .

[0109] Step d: Calculate the difference Δt” = t - t between the current timestamp and the start time of the yellow light y , calculate the time t s ” = Δt” % T when the current traffic light is in the starting state of the yellow light cycle.

[0110] Determine t s ” is less than the yellow light cycle T y If t s ” < T y , then the light color of the current traffic light is yellow, and the remaining time is t” = T y -t s ” ; If t s ” ≥ T y , then determine t s ” is less than T y +T r If t s ” < T y +T r , then the light color of the current traffic light is red, and the remaining time is t” = T y +T r -t s ; If t s ≥ T y +T r , then the light color of the current traffic light is green, and the remaining time is t” = T - t s ”.

[0111] If the countdown information of the current traffic light is recognized and there is cycle data of the current traffic light in the traffic light cycle table, after deriving the remaining time of the current traffic light, calculate the difference between the recognized countdown information of the current traffic light and the derived light color and countdown information of the current traffic light. If the difference between the two is large, it is considered that the cycle table data corresponding to the UUID of the current traffic light is invalid, and the cycle analysis is re-executed; if the difference between the two is small, the data derived is used to update the traffic light cycle table.

[0112] When retrieving according to the UUID of the current traffic light in the traffic light cycle table, if there is no cycle data of the current traffic light in the traffic light cycle table, a traffic light state speculation method based on Bayesian estimation is used to predict the switching probability of the light color of the current traffic light, as Figure 4 shown, including the following steps:

[0113] Step A: Obtain the average value and standard deviation of the duration of each light color of other traffic lights on the road where the current traffic light is located.

[0114] Step B: Assume that each light color of the traffic light follows a normal distribution, and calculate the likelihood function P(t s | state unchanged):

[0115]

[0116] In the formula, t s is the duration of the light color; σs is the standard deviation of the lamp color duration; T s is the average value of the lamp color duration.

[0117] Step C: Calculate the switching probability P (state about to change | t s ) of the current traffic light's lamp color s when it has lasted for t s :

[0118]

[0119] P(t s | state about to change) = 1 - P(t s | state unchanged);

[0120]

[0121] P(state unchanged) = 1 - P(state about to change);

[0122] P(t s ) = P(t s | state unchanged)P(state unchanged) + P(t s | state about to change)P(state about to change);

[0123] In the above formula, P(t s | state about to change) is the probability of observing a countdown of t s under the condition that the state is about to change, that is, the occurrence of a state change is associated with a specific time point; P(state about to change) is the probability that the state is about to change unconditionally, that is, the possibility of the traffic light state changing at any time; P(t s | state unchanged) is the probability of a countdown of t s on the premise that the traffic light state is known to remain unchanged, that is, the possibility of the countdown t s appearing under the current traffic light state; P(state unchanged) is the probability that the state remains in the current state without changing unconditionally, that is, the possibility of the traffic light state not changing at any time.

[0124] Step S4: Dynamically generate an optimal passing strategy based on the current traffic light's lamp color, remaining time, lamp color switching probability, vehicle's current speed, road speed limit, and the state of the vehicle in front.

[0125] Specifically, after obtaining the current traffic light's lamp color and countdown information through the above method, design the vehicle's passing strategy based on the traffic light countdown probability inference, as Figure 5 shown, including the following steps:

[0126] Step S41: Determine whether the current traffic light color is green. If so, execute Step S42; otherwise, execute Step S49.

[0127] Step S42: Determine whether countdown information is recognized. If so, execute Step S43; otherwise, execute Step S46.

[0128] Step S43: Calculate the minimum passing speed v of the vehicle min :

[0129]

[0130] where d is the distance of the vehicle from the stop line of the intersection; t is the remaining time of the green light.

[0131] Step S44: Determine whether the minimum passing speed is less than the set road speed limit v max , if v min <v max then execute Step S45; otherwise, execute Step S49.

[0132] Step S45: If there is a vehicle in front of the vehicle and the speed of the vehicle in front is greater than the minimum passing speed v min , then the vehicle passes at a speed of v min <v<v max ; otherwise, execute Step S49.

[0133] Step S46: If the probability P (state about to change|t s ) that the predicted traffic light color is about to turn green is less than the set probability threshold, then execute Step S49; otherwise, execute Step S47. The value range of the probability threshold is 0 to 1. The closer the probability threshold is to 1, the more aggressive the driving style of the vehicle is, and the system adjusts the vehicle to accelerate according to the calculated probability of traffic light change. The closer the probability threshold is to 0, the more conservative the driving style of the vehicle is, and the system controls the vehicle to decelerate relatively to avoid running a red light.

[0134] Step S47: If there is a vehicle in front of the vehicle and the speed of the vehicle in front is less than the road speed limit, then the vehicle passes at the speed of the vehicle in front; otherwise, it passes at the road speed limit.

[0135] Step S48: Determine whether the vehicle has passed through the intersection. If so, end the judgment; otherwise, return to Step S41 to re-judge.

[0136] Step S49: Decelerate and stop, return to Step S41, and wait for the traffic light color of the current traffic light to turn green.

[0137] Such as Figure 6As shown in the figure, the hardware supporting the embodiments of the present invention includes a long - focal - length camera 1 and a short - focal - length camera 2 arranged inside the windshield of the vehicle 5, which are connected to the domain controller 4 inside the vehicle through a camera signal line 3. The algorithms and logics of the above - mentioned method are all carried out in the domain controller 4.

[0138] The method provided by the embodiments of the present invention assigns a unique UUID to the traffic light through positioning information, and establishes a traffic - light cycle table by combining real - time perception, which can cope with the result jitter caused by the occlusion of the traffic light or unstable perception when passing the same traffic light later; at the same time, when the cycle table is not established, the probability of the traffic light about to change can also be given through the statistical values of the traffic lights on the same road section; by integrating the state of the vehicle in front, the distance from the stop line at the intersection, and speed - limit regulations, the method of the present invention can obtain a safer and more efficient intersection decision - making strategy; by combining the countdown and Bayesian inference, an efficient and safe traffic - passing decision - making scheme is provided for autonomous vehicles, which is applicable to various traffic - light scenarios.

[0139] The embodiments of the present invention also provide a driving adjustment system for autonomous vehicles based on countdown probability, which is implemented based on the above - mentioned method for adjusting the driving of autonomous vehicles based on countdown probability, and includes: a traffic - light detection and status recognition module, a traffic - light cycle - table construction and update module, a Bayesian probability inference module, a traffic - passing strategy calculation module, a vehicle control module, a data storage and feedback module, and a human - machine interaction module.

[0140] Traffic - light detection and status recognition module: Real - time detect the traffic - light status through a multi - focal - length camera, and identify the countdown information of the traffic light through a target - detection model and OCR technology;

[0141] Traffic - light cycle - table construction and update module: Construct a traffic - light cycle table by recording the timestamps of traffic - light status changes, the duration of each light color, and the start timestamp multiple times, and dynamically update the cycle table according to the real - time detection results;

[0142] Bayesian probability inference module: Assume that the duration of the light color follows a normal distribution, and calculate the probability of the traffic - light status remaining or changing based on the historical data in the traffic - light cycle table and the time that the current light color has lasted;

[0143] Traffic - passing strategy calculation module: Dynamically generate an optimal traffic - passing strategy according to the light color of the current traffic light, the countdown information, the current position, speed, speed limit of the vehicle, and the status of the vehicle in front;

[0144] Vehicle control module: Adjust the vehicle speed in real time according to the output of the traffic - passing strategy calculation module.

[0145] Data storage and feedback module: Used to store the detection results of traffic lights, traffic - light cycle - table data, Bayesian probability inference results, and the execution situation of traffic - passing strategies, dynamically update the traffic - light cycle - table data, and optimize the traffic - passing strategies.

[0146] Human-computer interaction module: Displays the current traffic light status, countdown information, traffic strategies, and system operating status to the driver, and provides voice or visual prompts for the driver.

[0147] The technical features of the above embodiments can be combined arbitrarily. For the sake of concise description, not all possible combinations of the technical features in the above embodiments are described. Only the preferred embodiments of the present invention are expressed. The description is relatively specific and detailed, but it should not be construed as a limitation on the scope of the present invention. As long as there is no contradiction in the combination of these technical features, it should be considered as within the scope described in this specification.

[0148] It should be noted that for those of ordinary skill in the art, without departing from the concept of the present invention, several modifications and improvements can still be made, and these all belong to the protection scope of the present invention. Therefore, the protection scope of the present invention should be subject to the appended claims.

Claims

1. A driving adjustment method for an autonomous vehicle based on countdown probability, characterized in that, It includes the following steps: Step S1: Construct a traffic light cycle table based on historical cycle data, identify the light color and countdown information of the current traffic light. If the historical cycle data of the current traffic light is recorded in the traffic light cycle table, calculate the remaining time of the light color of the current traffic light according to the historical cycle data; otherwise, execute Step S2; Step S2: Collect the cycle data of the current traffic light and update the traffic light cycle table. If the countdown information of the current traffic light is not identified in Step S1, execute Step S3; otherwise, execute Step S4; Step S3: Use the historical cycle data of other traffic lights in the section where the current traffic light is located to predict the switching probability of the light color of the current traffic light; Step S4: Dynamically generate an optimal passing strategy according to the light color, remaining time, light color switching probability, current vehicle speed, road speed limit and the state of the vehicle ahead.

2. The driving adjustment method of an autonomous vehicle based on countdown probability according to claim 1, characterized in that: In Step S1, a universally unique identifier UUID is assigned to each traffic light in the vehicle map according to the geographical location of the traffic light, and a traffic light cycle table is constructed based on the UUID, geographical location, duration of each light color, start time of each light color and the timestamp record of light color switching.

3. A driving adjustment method for an autonomous vehicle based on countdown probability according to claim 1, characterized in that: In Step S1, the light color and countdown information of the current traffic light are identified by a multi-focus camera, including the following steps: judge the distance between the vehicle and the traffic light. When the distance between the vehicle and the traffic light is greater than the set threshold, identify the light color and countdown information of the current traffic light through a long-focus camera; otherwise, identify the light color and countdown information of the current traffic light through a short-focus camera.

4. The driving adjustment method of an autonomous vehicle based on countdown probability according to claim 1, characterized in that: The step of calculating the remaining time of the light color of the current traffic light according to the historical cycle data in Step S1 includes the following steps: Step S11: Obtain the duration T of the yellow light, the start time t of the yellow light, the duration T of the red light, the start time t of the red light, the duration T of the green light, and the start time t of the green light in the historical cycle data from the traffic light cycle table according to the UUID of the current traffic light, and calculate the complete cycle T of the traffic light as T = T + T + T. If the current traffic light color is green, execute Step S12; if it is red, execute Step S13; otherwise, execute Step S14; y and the start time t of the yellow light y the duration T of the red light r and the start time t of the red light r the duration T of the green light g and the start time t of the green light g , calculate the complete cycle T of the traffic light as T = T y + T r + T g . If the current traffic light color is green, execute Step S12; if it is red, execute Step S13; otherwise, execute Step S14; Step S12: Calculate the difference Δt = t - t between the current timestamp and the start time of the green light g , calculate the time t when the current traffic light is in the starting state of the green light cycle s = Δt % T; Determine if t s is less than the duration T of the green light g , if t s < T g , then the color of the current traffic light is green, and the remaining time is t = T g - t s ; if t s ≥ T g , then determine if t s is less than T g + T y , if t s < T g + T y , then the color of the current traffic light is yellow, and the remaining time is t = T g + T y - t s ; if t s ≥ T g + T y , then the color of the current traffic light is red, and the remaining time is t = T - t s ; Step S13: Calculate the difference Δt' = t - t between the current timestamp and the start time of the red light r , calculate the time t s ' when the current traffic light is in the starting state of the red light cycle; determine if t s ' is less than the red light cycle T r , if t s ' < T r , then the light color of the current traffic light is red, and the remaining time is t' = T r - t s '; if t s ' ≥ T r , then determine if t s ' is less than T g + T r , if t s ' < T g + T r , then the light color of the current traffic light is green, and the remaining time is t' = T g + T r - t s ; if t s ≥ T g + T r , then the light color of the current traffic light is yellow, and the remaining time is t' = T - t s '; Step S14: Calculate the difference Δt” = t - t between the current timestamp and the start time of the yellow light, and calculate the time t” when the current traffic light is in the starting state of the yellow light cycle, where t” = Δt” % T; determine whether t” is less than the yellow light cycle T. If t” < T, then the color of the current traffic light is yellow, and the remaining time is t” = T - t”. If t” ≥ T, then determine whether t” is less than T + T. If t” < T + T, then the color of the current traffic light is red, and the remaining time is t” = T + T - t”. If t” ≥ T + T, then the color of the current traffic light is green, and the remaining time is t” = T - t”. y Calculate the time t when the current traffic light is in the starting state of the yellow light cycle s ” = Δt” % T; Determine t s ” Whether it is less than the yellow light cycle T y If t s ” < T y Then the color of the current traffic light is yellow, and the remaining time is t” = T y - t s ” If t s ” ≥ T y Then determine whether t s ” is less than T y + T r If t s ” < T y + T r Then the color of the current traffic light is red, and the remaining time is t” = T y + T r - t s If t s ≥ T y + T r Then the color of the current traffic light is green, and the remaining time is t” = T - t s ”.

5. A driving adjustment method for an autonomous vehicle based on countdown probability according to claim 1, characterized in that: The step of collecting the cycle data of the current traffic light and updating the traffic light cycle table in Step S2 includes the following steps: Step S21: Obtain the UUID of the current traffic light, judge the light color switching state of the current traffic light. If the current traffic light changes from green to yellow, execute Step S22; if the current traffic light changes from yellow to red, execute Step S23; if the current traffic light changes from red to green, execute Step S24; Step S22: Record the timestamp of the start time of the yellow light, the timestamp when the yellow light changes to red and the timestamp when the red light changes to green, calculate the cycles of the yellow light and the red light, and execute Step S25; Step S23: Record the timestamp of the start time of the red light, the timestamp when the red light changes to green, calculate the cycle of the red light, and execute Step S25; Step S24: Record the timestamp of the start time of the green light; Step S25: Judge whether the vehicle is passing the current traffic light for the second time according to the UUID of the current traffic light. If so, calculate the difference between the start timestamps of the two same-color lights recorded, divide the difference of the timestamps by n, judge whether the remainder is close to 0, and whether the quotient is greater than the sum of the cycles of the yellow light and the red light. If so, update the traffic light cycle table; otherwise, let n=n + 1; Step S26: Repeat Step S25 until the remainder of the difference between the timestamps of the start times of all light colors divided by n is close to 0, and the quotient is greater than the sum of the cycles of the yellow light and the red light; Step S27: Update the traffic light cycle table according to the UUID, geographical location, duration of each light color, start time of each light color, and timestamp of light color switching of the current traffic light.

6. A driving adjustment method for an autonomous vehicle based on countdown probability according to claim 1, characterized in that: Step S3 includes the following steps: Step S31: Obtain the average value and standard deviation of the duration of each light color of other traffic lights on the section where the current traffic light is located; Step S32: Assume that the colors of the traffic lights follow a normal distribution, and calculate the likelihood function P(t s | the state remains unchanged) of the current traffic light color s remaining unchanged: where t s is the duration of the lamp color; σ s is the standard deviation of the duration of the lamp color; T s is the average value of the duration of the lamp color. Step S33: Calculate the probability P (state about to change|t s ) that the current traffic light color s has lasted for t s when the current traffic light color is about to change: P(t s | about to change state) = 1 - P(t s | state unchanged); P(Status unchanged)=1 - P(Status about to change); P(t s ) = P(t s | unchanged state)P(unchanged state) + P(t s | about to change state)P(about to change state); In the above formula, P(t s | the state is about to change) is the probability of observing a countdown of t s under the condition that the state is about to change; P(the state is about to change) is the probability that the state is about to change unconditionally; P(t s | the state remains unchanged) is the probability of a countdown of t s given that the traffic light state remains unchanged; P(the state remains unchanged) is the probability that the state remains the current state without changing unconditionally.

7. A driving adjustment method for an autonomous vehicle based on countdown probability according to claim 1, characterized in that: Step S4 includes the following steps: Step S41: Determine whether the light color of the current traffic light is green. If so, execute Step S42; otherwise, execute Step S49; Step S42: Determine whether countdown information is recognized. If so, execute Step S43; otherwise, execute Step S46; Step S43: Calculate the minimum passing speed v of the vehicle min : In the formula, d is the distance of the vehicle from the stop line of the intersection; t is the remaining time of the green light; Step S44: Determine whether the minimum passing speed is less than the road speed limit v max If v min < v max then execute Step S45; otherwise, execute Step S49; Step S45: If there is a vehicle in front of the vehicle and the speed of the vehicle in front is greater than the minimum passing speed v min , then the vehicle passes at a speed of v min < v < v max , otherwise, execute Step S49; Step S46: If the probability that the predicted light color is about to turn green is less than the set probability threshold, execute Step S49; otherwise, execute Step S47; Step S47: If there is a vehicle in front of the vehicle and the speed of the vehicle in front is less than the road speed limit, the vehicle passes at the speed of the vehicle in front; otherwise, it passes at the road speed limit; Step S48: Determine whether the vehicle has passed the intersection. If so, end the judgment; otherwise, return to Step S41 for re-judgment; Step S49: Decelerate and stop, return to Step S41, and wait for the light color of the current traffic light to turn green.

8. A driving adjustment system for an autonomous vehicle based on countdown probability, implemented based on a driving adjustment method for an autonomous vehicle based on countdown probability according to any one of claims 1 to 7, characterized in that Includes: Traffic light detection and status recognition module, traffic light cycle table construction and update module, Bayesian probability calculation module, traffic strategy calculation module, and vehicle control module; The traffic light detection and status recognition module: Real-time detect the traffic light status through a multi-focal length camera, and recognize the countdown information of the traffic light through a target detection model and OCR technology; The traffic light cycle table construction and update module: Construct a traffic light cycle table by recording the timestamps of traffic light status changes, the duration of each light color, and the start time stamps multiple times, and dynamically update the traffic light cycle table according to the real-time detection results; The Bayesian probability calculation module: Assume that the light color duration follows a normal distribution, and calculate the probability of the traffic light status remaining or changing based on the historical data in the traffic light cycle table and the duration of the current light color; The traffic strategy calculation module: Dynamically generate an optimal traffic strategy according to the light color of the current traffic light, countdown information, current position, speed, speed limit of the vehicle, and the status of the vehicle in front; The vehicle control module: Adjust the vehicle speed in real time according to the output of the traffic strategy calculation module.

9. The driving adjustment system of an autonomous vehicle based on countdown probability according to claim 8, wherein: The system further includes a data storage and feedback module, which is used to store the detection results of traffic lights, traffic light cycle table data, Bayesian probability calculation results, and traffic strategy execution conditions, dynamically update the traffic light cycle table data, and optimize the traffic strategy.

10. The driving adjustment system of an autonomous vehicle based on countdown probability according to claim 9, wherein: The system further includes a human-machine interaction module, which displays the current traffic light status, countdown information, traffic strategy, and system operation status to the driver, and provides voice or visual prompts for the driver.

Images