Automatic pre-warning steering method and system

Abstract

The present application relates to the technical field of vehicle intelligent control, and particularly relates to an automatic early warning steering method and system, comprising: pre-collecting multiple sample images on a pre-travel path of a vehicle hub, generating a training data set based on the multiple sample images; constructing an identification network architecture, training the identification network architecture using the training data set to obtain a micro target identification network model; collecting real-time images on the pre-travel path of the vehicle hub in real time, extracting features of the real-time images using the micro target identification network model to obtain target information and target position information corresponding to the real-time images; processing the target information and the target position information to obtain a steering angle, controlling the vehicle to steer based on the steering angle to change the travel path of the vehicle hub. The present application is at least beneficial to realize detailed detection and early warning steering of the pre-travel path of the vehicle hub.

Description

TECHNICAL FIELD

[0001] The present application belongs to the technical field of vehicle intelligent control, and particularly relates to an automatic early warning steering method and system. BACKGROUND

[0002] With the gradual popularization of intelligent driving, more and more vehicle intelligent control strategies emerge in an endless stream, which are all aimed at preventing collision, pedestrian detection, dangerous object detection, etc., but are all large target-oriented intelligent control technologies. Moreover, the ordinary dynamic ranging technology, image-based target recognition technology and radar-based target recognition technology involved in the existing intelligent control technology all have a pain point, that is, slow processing speed, which still brings certain safety risks.

[0003] The existing intelligent control technology specifically includes automatic emergency braking technology, lane departure warning technology, blind spot monitoring technology, traffic signal recognition technology, driver monitoring technology, rear traffic crossing prompt technology and vehicle door opening warning technology. The automatic emergency braking technology uses millimeter wave radar for ranging, which can effectively measure the target within 170 meters in front of the vehicle. In adaptive cruise and normal driving, the vehicle calculates the emergency braking distance according to the front vehicle state, vehicle speed and acceleration state, and if the critical braking distance is approached, the vehicle will automatically emergency brake, thereby preventing accidents and improving driving safety. The lane departure warning technology mainly uses a camera to detect the lane auxiliary line for lane departure warning. This technology is developed and implemented using a traditional algorithm (non-deep learning algorithm). The blind spot detection technology, traffic signal recognition technology, driver monitoring technology, rear traffic crossing prompt technology and vehicle door opening warning technology all use a camera equipped with a target recognition intelligent algorithm. Different training sets are used for training to achieve the application purpose in different scenes.

[0004] However, the existing technology does not have an intelligent control technology for detailed detection of the pre-driving area of the vehicle hub. SUMMARY

[0005] Therefore, the present application aims to provide an automatic early warning steering method and system, which at least has the advantages of realizing detailed detection of the pre-driving path of the vehicle hub and early warning steering.

[0006] To achieve the above-mentioned purpose, the technical solution of the present application is as follows:

[0007] The application provides an automatic early warning turning method, which comprises the following steps: pre-collecting a plurality of sample images on a pre-travel path of a vehicle hub, generating a training data set based on the plurality of sample images; constructing an identification network architecture, training the identification network architecture by using the training data set, and obtaining a micro target identification network model; collecting a real-time image on the pre-travel path of the vehicle hub in real time, performing feature extraction on the real-time image by using the micro target identification network model, and obtaining target information and target position information corresponding to the real-time image; processing the target information and the target position information to obtain a turning angle, and controlling the vehicle to turn based on the turning angle, so as to change the travel path of the vehicle hub.

[0008] Further, the generation of the training data set based on the plurality of sample images comprises the following steps: pre-collecting N sample images on the pre-travel path of the vehicle hub, selecting M sample images from the N sample images for manual target labeling, obtaining labeled M sample images, training an identification model based on the labeled M sample images, labeling the remaining (N-M) sample images by using the identification model, and obtaining labeled N sample images.

[0009] Further, the establishment of the training data set based on the labeled N sample images comprises the following steps: performing image transformation on the labeled N sample images to expand the labeled N sample images into xN sample images, and establishing a training data set comprising the xN sample images, wherein x is an integer greater than 1.

[0010] Further, the micro target identification network model is a deep convolutional neural network model, and the deep convolutional neural network model comprises a first convolutional layer, a first pooling layer, a second convolutional layer, a second pooling layer, a third convolutional layer, a third pooling layer and a fully connected layer.

[0011] Further, the training of the identification network architecture by using the training data set to obtain the micro target identification network model comprises the following steps: inputting a sample image in the training data set into the identification network architecture, constructing a loss function, adjusting model parameters of the identification network architecture by using an optimization algorithm, and obtaining the micro target identification network model until the loss function is minimized.

[0012] Further, the loss function is as follows:

[0013]

[0014] wherein loss represents the loss function, y belongs to {1, 0}, y represents a target real class, p belongs to [0, 1], p represents an estimated probability when y = 1, (1-p) represents an estimated probability when y = 0, γ represents an adjustable static factor, β represents an adjustable dynamic factor, e p represents a probability amplification method.

[0015] Further, the target position information comprises lateral position information and distance information, and the target information comprises a target category; processing the target information and the target position information to obtain the steering angle comprises: calculating a steering coefficient according to the vehicle speed, the vehicle stability state, the distance information, the target category and the lateral position information by using a steering coefficient formula, and obtaining the steering angle by using the steering coefficient.

[0016] Further, the steering coefficient formula is as follows:

[0017]

[0018] wherein K represents the steering coefficient, V represents the current vehicle speed, V max represents the maximum design speed of the vehicle, S represents the stability state score of the vehicle, S max represents the maximum value of the stability state score of the vehicle, D represents the distance between the target and the vehicle, D min represents the minimum distance triggering the early warning, C represents the target category coefficient, C max represents the maximum value of the target category coefficient, P represents the lateral position coefficient, P max represents the maximum value of the lateral position coefficient, and a, b, c, d and e are all weight factors.

[0019] Further, obtaining the steering angle by using the steering coefficient comprises: calculating the steering angle by using a steering angle formula, and the steering angle formula is as follows:

[0020] θ=K×t 2 ;

[0021] wherein K represents the steering coefficient, and t represents the remaining steering time.

[0022] The application further provides another aspect of an automatic early warning steering system, which comprises: a data capture layer, the data capture layer being used to collect real-time images on a pre-travel path of a vehicle hub; a strategy customization layer, the strategy customization layer comprising a micro target recognition network model, the micro target recognition network model being used to extract features of the real-time images to obtain target information and target position information corresponding to the real-time images; and a real-time operation layer, the real-time operation layer being used to process the target information and the target position information to obtain a steering angle, and control the vehicle to steer based on the steering angle to change the travel path of the vehicle hub.

[0023] Compared with the prior art, the application can achieve the following beneficial effects: the embodiment of the application uses a camera to take real-time images of the pre-travel path of the vehicle hub in real time, and then uses a micro target recognition network model with high processing speed to process the real-time images, to identify micro dangerous objects on the pre-travel path of the vehicle hub in real time, and change the travel path of the vehicle hub according to the identification result, to prevent the vehicle from having a tire burst during travel, and further avoid personal injury caused by a tire burst. The significance of the automatic early warning steering method provided by the embodiment of the application lies in that during high-speed travel of the vehicle, the driver will not observe the small targets on the road in front in detail. Based on this, the embodiment of the application provides an automatic early warning steering method, which can realize real-time monitoring of the path to be passed by the wheel, and once an object that may pose a threat to the wheel is detected, the early warning mechanism is triggered to automatically adjust the travel direction of the vehicle, so as to prevent potential dangerous situations and ensure the safety of passengers and pedestrians. BRIEF DESCRIPTION OF DRAWINGS

[0024] The accompanying drawings, which form a part of the present application, are included to provide a further understanding of the application and are incorporated herein for explanation by illustrating a preferred embodiment of the present application. The present application will also be described in greater detail by way of specific examples, which illustrate preferred embodiments of the present application. In the drawings:

[0025] Figure 1 Network structure diagram of the micro target recognition network model described in the embodiment of the present application;

[0026] Figure 2 Use scenario diagram of the automatic early warning steering method described in the embodiment of the present application. DETAILED DESCRIPTION

[0027] In order to make the purpose, technical scheme and advantages of the present application clearer, the present application will be further described in detail below with reference to the drawings and specific embodiments. It should be understood that the specific embodiments described herein are only used to explain the present application and do not limit the present application.

[0028] It should be noted that the embodiments in the present application and the features in the embodiments can be combined with each other without conflict.

[0029] In the description of the present application, it should be understood that the terms "center", "longitudinal", "transverse", "length", "width", "thickness", "upper", "lower", "front", "rear", "left", "right", "vertical", "horizontal", "top", "bottom", "inner", "outer", "clockwise", "counterclockwise" and the like indicate the orientation or positional relationship based on the orientation or positional relationship shown in the drawings, and are only for the convenience of describing the present application and simplifying the description, and do not indicate or imply that the devices or elements referred to must have a particular orientation, be constructed and operated in a particular orientation, and therefore cannot be understood as limiting the present application. In addition, the terms "first", "second" and the like are only for descriptive purposes and cannot be understood as indicating or implying relative importance or implicitly indicating the number of technical features indicated. Therefore, the features defined with "first", "second" and the like can be explicitly or implicitly included one or more. In the description of the present application, unless otherwise stated, the meaning of "a plurality of" is two or more.

[0030] In the description of the present application, it should be noted that unless otherwise explicitly specified and limited, the terms "mounting", "connecting", "connecting" should be understood broadly, for example, it can be fixedly connected, or it can be detachably connected, or integrally connected; it can be mechanically connected, or it can be electrically connected; it can be directly connected, or it can be indirectly connected through an intermediate medium, or it can be connected inside two elements. For those skilled in the art, the specific meaning of the above terms in the present application can be understood through specific circumstances.

[0031] The present application will be described in detail below with reference to the accompanying drawings and in conjunction with the embodiments.

[0032] The present application provides an automatic early warning turning method, comprising: pre-collecting a plurality of sample images on a pre-travel path of a vehicle hub, generating a training data set based on the plurality of sample images; constructing a recognition network architecture, training the recognition network architecture using the training data set to obtain a micro target recognition network model; real-time collecting real-time images on the pre-travel path of the vehicle hub, using the micro target recognition network model to extract features of the real-time images to obtain target information and target position information corresponding to the real-time images; processing the target information and the target position information to obtain a turning angle, controlling the vehicle to turn based on the turning angle to change the travel path of the vehicle hub.

[0033] It should be noted that the target involved in the embodiments of the present application is a smaller dangerous object, which can be a nail, glass, blade, stone or brick, etc.

[0034] In some embodiments of the present application, a camera is used to collect sample images and real-time images. In some examples, the camera can be a visible light camera. In some examples, the camera can be placed in one of the following two positions: the position of the two front headlights of the vehicle or the position of the rearview mirror of the vehicle. In other examples, the camera can be placed in other positions on the front end of the vehicle as long as the camera can collect images of the pre-travel path of the vehicle hub.

[0035] In some embodiments of the present application, the number of cameras is usually two, for example, the first camera and the second camera shown in FIG. 1. Figure 2

[0036] In some embodiments of the present application, the camera can be used to collect images at a distance of 19-20 m in front of the vehicle hub.

[0037] Further, generating the training data set based on the plurality of sample images comprises: pre-collecting N sample images on the pre-travel path of the vehicle hub, selecting M sample images from the N sample images for manual target labeling, i.e., manually labeling smaller dangerous objects in the sample images to obtain labeled M sample images; training an identification model based on the labeled M sample images, labeling the remaining (N-M) sample images using the identification model to obtain labeled N sample images; and establishing a training data set according to the labeled N sample images.

[0038] In some embodiments of the present application, the specific steps of pre-collecting N sample images on the pre-travel path of the vehicle hub can be as follows: using a visible light camera to take real-time images of the pre-travel path in front of the vehicle hub, i.e., sample images. Before taking the images, various smaller dangerous objects can be placed on the pre-travel path in front of the vehicle hub, and then the sample images can be collected and stored. In some examples, 1000 sample image data can be collected, i.e., N=1000. Then, the sample images can be manually labeled in offline target labeling, and 100 sample images can be labeled, i.e., M=100. Then, the identification model can be trained using the data set of the 100 manually labeled sample images, and the remaining 900 sample images can be labeled using the trained identification model. After merging the 100 manually labeled sample images and the 900 sample images labeled by the identification model, a data set of 1000 labeled sample images is obtained. Each labeled sample image corresponds to corresponding image data, and the 1000 sample images and the corresponding image data constitute a training data set.

[0039] ​Further, the training data set is established according to the N sample images, and the N sample images are subjected to image transformation to expand the N sample images into xN sample images, and the training data set including the xN sample images is established, wherein x is an integer greater than 1.

[0040] In some embodiments of the present application, in order to increase the diversity and scale of the training data set, the N sample images are subjected to image transformation by using image transformation techniques, which can include horizontal mirror inversion, position translation adjustment, angle rotation transformation, region cropping processing and image compression operation. Through these image transformation operations, the capacity of the training data set can be expanded to 8 times of the original capacity, so as to generate a final training data set which is more abundant and more suitable for model training.

[0041] Further, referring to Figure 1 The obtained micro target recognition network model is a deep convolutional neural network model, which includes a first convolutional layer, a first pooling layer, a second convolutional layer, a second pooling layer, a third convolutional layer, a third pooling layer and a full connection layer. The deep convolutional neural network model uses convolution kernels of different sizes to extract features of different scales from sample images. In some examples, an image with a size of 640x640 pixels can be input into the deep convolutional neural network model, and the first convolutional layer can be used to perform a convolution operation with a step size of 2, a kernel size of 3x3 and 50 filters to obtain a feature map with a size of 320x320x50. The first pooling layer can be used to perform 2x2 pooling on the feature map output by the first convolutional layer to obtain a feature map with a size of 160x160x50. Then, the second convolutional layer can be used to perform a convolution operation with a step size of 2, a kernel size of 3x3 and 100 filters on the feature map output by the first pooling layer to obtain a feature map with a size of 81x81x100. The second pooling layer can be used to perform 2x2 pooling on the feature map output by the second convolutional layer to obtain a feature map with a size of 41x41x100. Then, the third convolutional layer can be used to perform a convolution operation with a step size of 2, a kernel size of 3x3 and 200 filters on the feature map output by the second pooling layer to obtain a feature map with a size of 21x21x200. The third pooling layer can be used to perform 2x2 pooling on the feature map output by the third convolutional layer to obtain a feature map with a size of 11x11x200. Finally, the full connection layer can be used to obtain a feature map with a size of 24200x100, i.e., 100 data, which include the coordinate positions (horizontal position information and distance information) of 20 targets, the lengths of the 20 targets, the widths of the 20 targets and the categories of the 20 targets.

[0042] The network architecture of the micro target recognition network model provided by the embodiment of the present application is simple, and compared with the existing YOLO (You Only Look Once) algorithm, the SSD (Single Shot MultiBox Detector) algorithm, the Faster RCNN algorithm and the NanoDet algorithm, the micro target recognition network model provided by the embodiment of the present application is more simple, and the micro target recognition network model provided by the embodiment of the present application can realize rapid detection of smaller targets, because only three features are extracted and full connection is performed once, the detection of smaller dangerous targets can be achieved.

[0043] Further, the training data set is used to train the recognition network architecture to obtain the micro target recognition network model, including: inputting the sample image in the training data set into the recognition network architecture, constructing a loss function, adjusting the model parameters of the recognition network architecture through an optimization algorithm until the loss function is minimum, so as to obtain the micro target recognition network model. That is, by constructing a deep feature learning framework (recognition network architecture), the deep feature learning framework uses diversified convolution filters to extract features from image data, and cooperatively learns by fusing these features, so as to generate a feature representation model for a specific target. Through repeated training iterations of the deep feature learning framework, the network parameters (weights) capable of tracking target features are optimized and determined, and the micro target recognition network model is obtained.

[0044] Further, the loss function is as follows:

[0045]

[0046] Wherein, loss represents the loss function, y∈{1, 0}, y represents the real class of the target, if the target is a real smaller dangerous article, y=1, if the target is a false smaller dangerous article, y=0, p∈[0, 1], p represents the estimated probability when y=1, (1-p) represents the estimated probability when y=0, γ represents an adjustable static factor, β represents an adjustable dynamic factor, e p represents a probability amplification method.

[0047] It should be noted that the loss function loss is a standard cross-entropy, which is represented by the product of -log(p t ) and a dynamic scaling factor (1-p t ) γ , and a probability weight weighted sum β(p t +e p ). The loss function provided by the embodiment of the present application increases the probability weight weighted sum β(p t +e p), which is conducive to increasing the detection accuracy of small dangerous targets. Through calculation, the average index of the average accuracy rate can be increased by 2 points.

[0048] The loss function loss is used to calculate the difference between the real target data and the training target data. The target data mainly includes the horizontal and vertical coordinates of the target coordinate frame (lateral position information and distance information), target length, target width and target category. After calculating the loss through the loss function, the weights in the model can be updated through the backpropagation gradient descent algorithm. Then, the collected real-time image is input into the trained micro target recognition network model to obtain target information. Then, the target position information can be calculated by combining the shooting distance of the camera lens. The target information and target position information are processed to obtain the steering angle, and the corresponding steering angle is transmitted to the control system of the vehicle to control the steering of the vehicle to avoid the smaller dangerous objects.

[0049] Further, the target position information includes lateral position information and distance information; the target information includes target category, and processing the target information and the target position information to obtain the steering angle includes: calculating a steering coefficient according to the vehicle speed, the vehicle stability state, the distance information, the target category and the lateral position information by using a steering coefficient formula to obtain the steering angle.

[0050] Further, the steering coefficient formula is as follows:

[0051]

[0052] wherein K represents the steering coefficient, V represents the current vehicle speed, the unit of the current vehicle speed is m / s or km / h, V max represents the maximum design speed of the vehicle, the maximum design speed and the current speed need to be in the same unit, S represents the stability state score of the vehicle, S is a dimensionless value between 0 and 1, 1 represents complete stability, S max represents the maximum value of the stability state score of the vehicle, S max = 1, D represents the distance (distance information) between the target and the vehicle, the unit of D can be m, D min represents the minimum distance for triggering the warning, the unit is m, C represents the target category coefficient, the target category coefficient C is a weight value given according to the target category, which can be a dimensionless value between 0 and 1, or a specific classification value, C max represents the maximum value of the target category coefficient, which can be 1 or the maximum weight value, P represents the lateral position coefficient, which is a weight value given according to the target lateral position information, which can be a dimensionless value between 0 and 1, or a specific classification value, P maxThe maximum value of the lateral position coefficient, which can be 1 or a maximum weight value, a, b, c, d, and e are weight factors for adjusting the degree of influence of each parameter on the steering coefficient, which can be optimized through experiments or simulations.

[0053] Further, the steering angle is obtained by the steering coefficient, including: using a steering angle formula to calculate the steering angle, and the steering angle formula is as follows:

[0054] θ=Kxt 2 ;

[0055] Wherein, K represents the steering coefficient, and t represents the remaining steering time. The remaining steering time refers to: if it is determined that the vehicle completes steering within 2 seconds, the remaining steering time from the start of steering to 2 seconds, for example, when starting steering, the remaining steering time is 2, and the steering angle required at this time is 4K. With the change of time, the remaining steering time gradually decreases, and the steering angle also gradually decreases.

[0056] In some embodiments of the application, K=5x10 -6 .

[0057] The automatic early warning steering method provided by the embodiments of the application can quickly and accurately detect small dangerous objects during vehicle driving, can realize real-time monitoring of the path to be passed by the vehicle wheel, and can trigger an early warning mechanism to automatically adjust the driving direction of the vehicle as soon as a dangerous object that may pose a threat to the vehicle wheel is detected, so as to prevent potential dangerous situations and ensure the safety of passengers and pedestrians.

[0058] The application further provides an automatic early warning steering system for implementing the automatic early warning steering method, the automatic early warning steering system comprising: a data capture layer, the data capture layer being configured to collect real-time images on a pre-driving path of a vehicle hub; a strategy customization layer, the strategy customization layer comprising a micro-target recognition network model, the micro-target recognition network model being configured to extract features of the real-time images to obtain target information and target position information corresponding to the real-time images; and a real-time operation layer, the real-time operation layer being configured to process the target information and the target position information to obtain a steering angle, and control steering of the vehicle based on the steering angle to change the driving path of the vehicle hub.

[0059] The automatic early warning steering system provided by the application comprises a data capture layer, a strategy customization layer, and a real-time operation layer, the data capture layer can comprise a visible light camera arranged at the front end of the vehicle, the strategy customization layer can be an intelligent processing module of the vehicle, and the real-time operation layer can be a steering control system of the vehicle.

[0060] It should be understood that the various forms of flow shown above can be used to reorder, add, or remove steps. For example, the steps recited in the present disclosure can be performed in parallel, in series, or in a different order, as long as the desired results of the present disclosure are achieved, which is not limited herein.

[0061] The specific implementation described above does not constitute a limitation on the protection scope of the present application. Those skilled in the art should understand that various modifications, combinations, sub-combinations, and substitutions can be made according to design requirements and other factors. Any modifications, equivalent replacements, and improvements made within the spirit and principles of the present application shall be included in the protection scope of the present application.

Claims

1. An automatic pre-warning turning method, characterized by, The method comprises the following steps: Pre-acquire a plurality of sample images on the pre-travel path of the vehicle hub, generate a training data set based on the plurality of sample images, which includes pre-acquiring N sample images on the pre-travel path of the vehicle hub, selecting M sample images from the N sample images for manual target labeling operation to obtain labeled M sample images; training an identification model based on the labeled M sample images, using the identification model to label the remaining (N-M) sample images to obtain labeled N sample images; establishing the training data set according to the labeled N sample images, N=1000, M=100; Construct an identification network architecture, train the identification network architecture using the training data set to obtain a micro target identification network model, which includes inputting the sample images in the training data set into the identification network architecture and constructing a loss function, adjusting the model parameters of the identification network architecture through an optimization algorithm until the loss function is minimized to obtain the micro target identification network model, the micro target identification network model is a deep convolutional neural network model, the deep convolutional neural network model includes a first convolutional layer, a first pooling layer, a second convolutional layer, a second pooling layer, a third convolutional layer, a third pooling layer, and a fully connected layer, the first convolutional layer performs a convolution operation on the image with a step size of 2, 3x3x50, the second convolutional layer performs a convolution operation on the feature map output by the first pooling layer with a step size of 2, 3x3x100, the third convolutional layer performs a convolution operation on the feature map output by the second pooling layer with a step size of 2, 3x3x200, the micro target identification network model is used to process real-time images to identify micro hazards on the pre-travel path of the vehicle hub in real time; the loss function is as follows: ； wherein, represents a loss function, , represents a target real class, , represents an estimated probability at time t, represents an estimated probability at time t, represents an adjustable static factor, represents an adjustable dynamic factor, represents a method of amplifying probabilities; Real-time acquisition of real-time images on the pre-travel path of the vehicle hub, feature extraction of the real-time images using the micro target identification network model to obtain target information and target position information corresponding to the real-time images, the target position information including lateral position information and distance information, the target information including target categories; Processing the target information and the target position information to obtain a steering angle, controlling the vehicle to turn based on the steering angle to change the travel path of the vehicle hub, processing the target information and the target position information to obtain a steering angle includes calculating a steering coefficient using a steering coefficient formula based on vehicle speed, vehicle stability state, distance information, target categories, and lateral position information to obtain the steering angle, the steering coefficient formula is as follows: ； Wherein, K represents a steering coefficient, V represents a current vehicle speed, Vmax represents a maximum design speed of the vehicle, S represents a stable state score of the vehicle, Smax represents a maximum value of the stable state score of the vehicle, D represents a distance between a target and the vehicle, Dmin represents a minimum distance triggering a warning, C represents a target category coefficient, Cmax represents a maximum value of the target category coefficient, P represents a lateral position coefficient, Pmax represents a maximum value of the lateral position coefficient, and a, b, c, d and e are all weight factors; The steering angle is obtained by the steering coefficient, and the steering angle is calculated by using a steering angle formula, and the steering angle formula is as follows: ； wherein K represents a turning coefficient, t represents a remaining turning duration, K .

2. The automatic pre-warning turning method according to claim 1, characterized in that, The training data set is established according to the N labeled sample images, and the N labeled sample images are subjected to image transformation to expand the N labeled sample images into xN sample images, and the training data set including the xN sample images is established, wherein x is an integer greater than 1.

3. An automatic warning turn system characterized by, The automatic warning steering system is used to implement the automatic warning steering method in any one of claims 1 to 2, and the automatic warning steering system comprises: A data capture layer is configured to collect real-time images on a pre-travel path of a vehicle hub; A strategy customization layer comprises a micro-target recognition network model configured to extract features of the real-time images to obtain target information and target position information corresponding to the real-time images; A real-time operation layer is configured to process the target information and the target position information to obtain a steering angle, and control the vehicle to steer based on the steering angle to change the travel path of the vehicle hub.

Images