A method for adaptive speed limit control of a curve of a road

Abstract

The application discloses a kind of self-adapting road curvature's bend speed limit control method, comprising the following steps: step 1: based on vehicle speed and driver characteristic index self-learning to obtain weight coefficient;Step 2: based on camera recognition lane line curvature;Step 3: optimal bend speed is obtained by using dynamic programming method.The application has the advantages that: it can improve the accuracy of vehicle driving on curve, vehicle stability and according to the driving preference of driver, adaptively according to different bend radius, different degree of speed decision for driving on curve is made.Driving speed planning is adaptively considered for bend, and vehicle state and driver characteristics are comprehensively optimized.

Description

TECHNICAL FIELD

[0001] The present application relates to the field of automatic driving, in particular to a bend speed limit control method adaptive to road curvature. BACKGROUND

[0002] In recent years, more and more automatic driving vehicles are driving on the road. At present, the open road in the city has complex driving scenes and changing road curvature. In order to improve the driving efficiency, the automatic driving vehicle needs to slow down when the road curvature is large, such as sharp turns, right-angle turns or U-turn scenes. Therefore, the automatic driving vehicle needs to slow down in advance when driving from a road with large bend radius to a road with small bend radius, so as to avoid accidents such as rollover caused by too fast speed due to untimely deceleration.

[0003] The bend curvature recognized by the existing technology through the monocular camera has certain random error and steady error, and sometimes there is certain missing recognition through the monocular camera recognition without high-precision positioning. In the case of being unable to recognize the bend curvature in advance, it will add certain uncertainty and safety hazards to the auxiliary driving.

[0004] The traditional bend speed planning usually only gives a pre-warning of the highest speed according to the lateral acceleration to ensure the safety of the vehicle through the bend, and lacks integrated planning of the bend driving speed according to the actual road geometric shape, vehicle dynamics and driver demand. Especially for different driving demands, the accuracy of vehicle bend driving, vehicle stability and driver's driving preference will all have different degrees of influence on the bend driving speed decision. Giving only the highest speed limit cannot realize the optimal automatic driving experience through the bend, so the existing technology of bend speed planning only gives a pre-warning limit of the highest speed, which cannot achieve the optimal bend passing. SUMMARY

[0005] The present application aims to overcome the shortcomings of the prior art and provide a bend speed limit control method adaptive to road curvature, which can realize the optimal bend speed matching the road curvature, vehicle state and driver characteristics.

[0006] In order to achieve the above-mentioned purpose, the technical scheme adopted by the present application is as follows: a bend speed limit control method adaptive to road curvature, comprising the following steps:

[0007] Step 1: self-learning the weight coefficient based on the vehicle speed and the driver characteristic index;

[0008] Step 2: recognizing the lane line curvature based on the camera;

[0009] Step 3: using the dynamic programming method to obtain the optimal bend speed.

[0010] The method comprises the following steps: in step 1, the acquired vehicle speed and driver characteristic index are subjected to fuzzy control to obtain a weight coefficient.

[0011] The fuzzy control comprises fuzzification, an inference engine and defuzzification; the vehicle speed and the driver characteristic index are taken as the input of the fuzzy control, are subjected to fuzzification first, then the inference engine forms a fuzzy conclusion based on a knowledge base, and the fuzzy conclusion is subjected to defuzzification to obtain the weight coefficient p.

[0012] The acquired driver characteristic index is a stable parameter based on historical driving data and reflects the driving tendency of the driver.

[0013] The fuzzy control is realized by using a pre-designed fuzzy controller, wherein the input of the fuzzy controller is the vehicle speed and the driver characteristic index, and the output is the weight coefficient.

[0014] The fuzzification can translate the input into a fuzzy set and establish a membership function; the defuzzification can interpret the fuzzy conclusion as an actual physical output, and the defuzzification is realized by using the COG gravity method.

[0015] In step 2, the lane line curvature is recognized based on a camera, and the bend speed limit value is obtained based on the lane line curvature.

[0016] The dynamic programming algorithm is used to realize the optimal bend speed meeting the road curvature, the vehicle state and the driver characteristic based on the bend speed limit value and the weight coefficient p.

[0017] The present application has the advantages that the accuracy of the vehicle bend driving can be improved, the vehicle stability can be improved, and the bend driving speed decision can be made according to the driving preference of the driver, the bend driving speed planning can be adaptively made according to different bend radii, and the road curvature, the vehicle state and the driver characteristic are comprehensively optimized. BRIEF DESCRIPTION OF DRAWINGS

[0018] The contents expressed by the drawings of the present application and the marks in the drawings are briefly described as follows:

[0019] Figure 1 It is a schematic diagram of the method principle of the present application.

[0020] Figure 2 It is a schematic diagram of the solution flow of the planning algorithm of the present application. DETAILED DESCRIPTION

[0021] The specific embodiments of the present application are further described in detail by comparing the drawings and describing the optimal embodiments.

[0022] The specific content of the present application is: the application mainly includes three parts, namely, learning different weight coefficients according to vehicle speed and driver characteristic index; the lane line curvature recognized by the camera; and the optimal solution of the dynamic programming method to obtain the optimal curved driving speed.

[0023] In the pretreatment, the vehicle speed and the driver characteristic index (a relatively stable parameter based on historical driving data to reflect the driving tendency of the driver) are first obtained; wherein the driver characteristic index refers to the driver characteristic index, which is a relatively stable parameter based on historical driving data to reflect the driving tendency of the driver. It includes the acceleration pedal opening, the brake pedal opening, the number of times and the time interval of the acceleration pedal opening, the number of times and the time interval of the brake pedal opening, and the number of overtaking in 0.5 hours at high speed. The characteristic quantization model of the driver is I=0, which represents that the driver characteristic is conservative; I=0.5, which represents that the driver characteristic is general. I=1, which represents that the driver characteristic is aggressive. The value range of I is 0-1.

[0024] Then the fuzzy controller is designed, which is a two-input single-output form, that is, the vehicle speed and the driver characteristic index are taken as the system input, and the weight coefficient is taken as the system output. Fuzzification can convert the input quantity into a fuzzy set and establish a membership function; the domain of the vehicle speed is [0, 150], and the domain of the driver characteristic index is [0, 3], which are converted into fuzzy subsets [NB, NM, NS, ZE, PS, PM, PB]. For the vehicle speed, NB represents the minimum speed, NM represents the slightly small speed, NS represents the relatively small speed, ZE represents the moderate speed, PS represents the relatively large speed, PM represents the slightly large speed, and PB represents the maximum speed. Similarly, for the driver characteristic index, NB represents the economical driving style, PB represents the aggressive driving style, ZE represents the general driving style, NM and NS are between the economical driving style and the general driving style, and PS and PM are between the general driving style and the aggressive driving style. The membership function is represented by a triangular function, and the calculation method is the barycentric method. In the fuzzy control reasoning process, the membership function adopts the "minimum" principle. The inference engine uses the rules and data in the knowledge base to generate fuzzy conclusions, and its basic tasks include two parts, one is to match the current input with which rules and data are related, and the other is to derive conclusions using the current input and reasoning; The rules of the controller are obtained based on expert knowledge and accumulated experience of skilled personnel, which is a logical representation of human perception reasoning. Defuzzification can interpret the fuzzy conclusion as an actual physical quantity output to be implemented by the actuator. The defuzzification adopts the COG barycentric method.

[0025] Then, the curved speed limit is obtained according to the lane line curvature recognized by the camera;

[0026] In order to get the real-time optimal speed limit of the curve, the dynamic programming method is used to get the optimal speed limit of the curve based on the dynamic programming algorithm. The theoretical basis of the algorithm is the optimality principle, which states that the optimal strategy for a multi-stage decision problem is also optimal for the state after the stage. According to this simple idea, dynamic programming converts a multi-stage decision problem into a series of single-step decision problems, i.e. first, from the last step state, recursively from each step decision to the initial stage state, and then reconstruct the optimal solution in the forward direction. The following three points are mainly considered: establishing a state equation reflecting the trend of dynamic system change; formulating a cost function to evaluate the merits of different strategies; constructing a constraint function set necessary for the system to operate within the normal range. For example, set the dynamic system composed of n stages, the state equation of the system can be expressed as: k+1 =T k (x k ,u k ),k=0,1,...n-1 Add the cost function, the system form is: Based on the current measurement information of the adaptive road curvature speed limit, including some kinematic parameters, the optimal output is obtained by optimization in a limited time, and the output is directly applied to the input index of the next moment, which is solved online. In a set time, the input and output parameters of the system are constantly refreshed until the required performance is reached. Through the derivation of the algorithm, the optimal speed of each step of the curve can be obtained by considering the road curvature, vehicle state and driver characteristics, etc.

[0027] The dynamic programming algorithm mainly considers the following three points: establishing a state equation reflecting the trend of dynamic system change; formulating a cost function to evaluate the merits of different strategies; constructing a constraint function set necessary for the system to operate within the normal range.

[0028] State equation:

[0029]

[0030] Where k = 0, 1, 2…N-1.

[0031] In the formula: v k and v k+1 are the vehicle speeds of k and k+1 stages based on the speed limit of the curve obtained by curvature; Δs is the length of the single planning road; i k is the gear ratio of k stage; δ k is the vehicle heading angle of k stage; T ek is the engine torque of k stage.

[0032] Constraint function:

[0033] According to the vehicle speed and the driver characteristic index, different weight coefficients and speed change rates are self-learned as constraint conditions

[0034]

[0035] Effects achieved: different weight coefficients are self-learned according to the driver characteristic index, and the speed change rate in a section of curve cannot exceed the comfortable change rate, so that the driving comfort is ensured in the whole process.

[0036] Obviously, the specific implementation of the present application is not limited by the above-mentioned manner, and various non-essential improvements made by adopting the method concept and technical solutions of the present application are within the protection scope of the present application.

Claims

1. A method for adaptive speed limit control on curved roads based on road curvature, the method comprising: The method comprises: obtaining a weight coefficient based on vehicle speed and driver characteristic index; identifying lane line curvature based on camera; obtaining optimal curve speed based on the obtained weight coefficient, lane line curvature and dynamic programming method; the method further comprises obtaining a weight coefficient based on fuzzy control of the obtained vehicle speed and driver characteristic index; the obtained driver characteristic index is a stable parameter based on historical driving data to reflect the driving tendency of the driver.

2. The method of claim 1, wherein: The fuzzy control comprises fuzzification, inference engine and defuzzification; the vehicle speed and driver characteristic index are taken as the input of the fuzzy control, first fuzzification is performed, then the inference engine is used to form a fuzzy conclusion based on the knowledge base, and the weight coefficient p is obtained by defuzzification of the fuzzy conclusion.

3. The method of claim 1, wherein: The fuzzy control is realized by using a pre-designed fuzzy controller, wherein the input of the fuzzy controller is the vehicle speed and the driver characteristic index, and the output is the weight coefficient.

4. The method of claim 2, wherein: Fuzzification can translate the input quantity into a fuzzy set and establish a membership function; defuzzification can interpret the fuzzy conclusion as an actual physical quantity output.

5. The method of claim 1, wherein: In step 2, the lane line curvature is identified based on the camera, and the curve speed limit value is obtained based on the lane line curvature.

6. The adaptive curve speed limit control method of claim 1-5, wherein: the dynamic programming algorithm is used to obtain the optimal curve speed that meets the road curvature, vehicle state and driver characteristics based on the curve speed limit value and the weight coefficient p.

7. The method of claim 2, wherein: The defuzzification is realized by using the COG gravity center method.

Images