An intelligent car motion control system based on visual recognition and autonomous navigation
By employing a multi-sensor fusion perception system combining dual-row grayscale sensors and color block sensors with a vision module on the intelligent vehicle, the problems of insufficient path tracking and dynamic environment adaptability in existing systems are solved. This achieves low-cost and efficient collaboration between visual recognition and motion control, improving path tracking accuracy and system integration.
Patent Information
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- SHANGHAI INST OF TECH
- Filing Date
- 2026-02-10
- Publication Date
- 2026-05-29
AI Technical Summary
Existing intelligent vehicle systems lack the ability to track paths and adapt to dynamic environments. In particular, systems based on a single sensor lack multimodal information fusion. High-performance processor solutions are costly and difficult to deploy in resource-constrained embedded systems. Separate perception and control architectures have low integration and poor real-time performance.
It adopts a dual-row, eight-channel grayscale sensor and color block sensor, combined with a vision module and a main control module to achieve multi-sensor fusion perception. Through the STM32F103C8T6 main control module and the OPENMV vision module, combined with the PID closed-loop control algorithm, it achieves efficient and collaborative visual recognition and motion control.
It achieves high-precision tracking of complex paths, ensuring that the vehicle can drive smoothly on straight lines, curves and complex paths. It has low-cost, real-time embedded control capabilities, which improves the system's integration and robustness.
Smart Images
Figure CN122108127A_ABST
Abstract
Description
Technical Field
[0001] This invention relates to the field of intelligent vehicle and mobile robot control technology, and in particular to an intelligent vehicle motion control system based on visual recognition and autonomous navigation. Background Technology
[0002] With the development of artificial intelligence and computer vision technologies, intelligent vehicles are attracting increasing attention as important automated execution units in scenarios such as educational experiments, indoor inspections, small-scale logistics transportation, and intelligent service robots.
[0003] Currently, existing intelligent vehicle systems based on microcontrollers or embedded platforms, both domestically and internationally, rely on single sensors for path tracking. For example, using infrared or grayscale sensors for simple line following lacks the fusion processing of multimodal information such as color and image data, and thus lacks adaptability to complex paths and dynamic environments. Visual navigation solutions based on high-performance processors, such as those using the NVIDIA Jetson platform for deep learning navigation, are costly and power-hungry, making them difficult to deploy in resource-constrained embedded systems. Separate perception and control architectures, with separate vision and motion control modules, result in low system integration, poor real-time performance, and difficulty in achieving multimodal task collaboration. Summary of the Invention
[0004] To address the shortcomings of existing technologies, this invention provides an intelligent vehicle motion control system based on visual recognition and autonomous navigation, aiming to achieve multi-sensor (visual, grayscale, color) fusion perception, low-cost embedded real-time control, and efficient collaboration between visual recognition and motion control.
[0005] To achieve the above objectives, the present invention provides an intelligent vehicle motion control system based on visual recognition and autonomous navigation, comprising: a main control module and a vision module, a pathfinding module, a drive module and a power supply module connected to the main control module; The pathfinding module includes eight grayscale sensors and color block sensors configured on the bottom of the intelligent vehicle; the eight grayscale sensors are divided into two rows along the direction of travel, with four sensors evenly distributed laterally in each row; the color block sensors are located between the two rows of grayscale sensors or in front of the two rows of grayscale sensors. The main control module collects the outputs of the pathfinding module and the vision module in real time, and the generated control signals are connected to the drive module to drive the motor of the intelligent car to run and control the driving speed and direction. The power module supplies power to the intelligent vehicle's motion control system.
[0006] Furthermore, it also includes a voice module and a remote control handle connected to the main control module; The main control module receives commands from the remote control handle and status feedback from the voice module to control the movement of the smart car.
[0007] Furthermore, the main control module adopts STM32F103C8T6, and the vision module adopts OPENMV.
[0008] This invention also provides a trajectory tracking method for an intelligent vehicle based on visual recognition and autonomous navigation. Based on the aforementioned intelligent vehicle motion control system based on visual recognition and autonomous navigation, the method includes the following steps: (1) Real-time acquisition of voltage values collected by 8 grayscale sensors and mapping them to grayscale values; (2) Based on the position of the grayscale sensor, the grayscale value of each row of grayscale sensors is weighted to obtain the weighted grayscale value of each row of grayscale sensors, and the lateral offset, yaw angle and path curvature of each row of grayscale sensors are calculated. (3) Construct a state machine based on the gray values obtained from the 8 grayscale sensors to identify the path features of the intelligent vehicle; (4) Based on the path characteristics, output control quantity based on the lateral offset of each row of grayscale sensors, as well as the yaw angle and path curvature.
[0009] Furthermore, step (1) specifically includes: in: g i For the first i The grayscale sensor collects the grayscale value corresponding to the voltage. V i For the first i The voltage value collected by the grayscale sensor. V min and V max These are the voltage reference values corresponding to the black and white lines, respectively.
[0010] Furthermore, step (2) specifically involves: Eight grayscale sensors are arranged in two rows along the travel path, with four sensors evenly distributed laterally in each row. Taking the center of the grayscale sensor array as the origin, the lateral coordinates of the eight grayscale sensors are: in: d The distance between adjacent grayscale sensors in each row. i =1...8, which are the numbers of the grayscale sensors, numbered from the front row to the back row; x i For the first i The lateral coordinates of the road grayscale sensor; Based on the location of each grayscale sensor, a Gaussian weight is applied to obtain the weight coefficient for each grayscale sensor: in: To control the width of the weight distribution; For the first i Weighting coefficients of the grayscale sensor; The normalized gray values of each row of grayscale sensors are weighted based on weighting coefficients to obtain the weighted gray values of each row: Based on the weighted gray values, the lateral offset, yaw angle, and path curvature of each row are obtained; in: The weighted grayscale values for the front row. Weighted grayscale values for the front row; This represents the lateral deviation of the front-row grayscale sensors. The lateral deviation of the rear grayscale sensor G 0 represents the theoretical grayscale value of the path center; K e This is the offset gain coefficient, used to convert grayscale differences into physical offsets; θ Yaw angle; L This refers to the distance between the two rows of grayscale sensors. K For path curvature.
[0011] Furthermore, step (3) specifically involves: (3.1) The gray values of the obtained 8 gray sensors are binarized according to the threshold. If the gray value is higher than or equal to the threshold, it is set to 1, indicating that it is in the white area. If it is lower than the threshold, it is set to 0, indicating that it is in the black area. The two rows of gray sensors obtain two sets of four-bit binary sequences, which form a spatial pattern vector. (3.2) The state machine predefines five typical path characteristics, including straight road state, starting to enter a curve state, turning state, crossroads state, and lost path state: (3.3) Based on the spatial pattern vector obtained within the time period T, perform state machine judgment to obtain the path features of the current intelligent vehicle.
[0012] Furthermore, if both sets of four-bit binary sequences in the acquired spatial pattern vector are
[0011] or
[1100] , and remain unchanged within the time period T, then the current path feature of the intelligent vehicle is a straight road state. If the second group of four-bit binary sequence in the acquired spatial pattern vector is still
[0011] or
[1100] , and the first group of four-bit binary sequence becomes
[0111] or
[1110] or
[0001] or
[1000] , and this lasts for a period of time T, then the current path characteristic of the intelligent vehicle is that it is starting to enter a curve state. If a set of four-bit binary sequences in the acquired spatial pattern vector shows discontinuous black / white signals or is distributed on both sides for a duration of T, then the current path characteristic of the intelligent vehicle is a crossroads state. If the second group of four-bit binary sequences in the acquired spatial pattern vector are both
[0000] or
[1111] and last for a period of time T, then the current path feature of the intelligent vehicle is a lost path state.
[0013] Furthermore, in the case of a straight road, the lateral offset of the front grayscale sensor is converted into a motor control command by the PID controller; at this time, the yaw angle and path curvature are not considered. in, u (t) is the control variable. K p , K i , K d These are the gains of the proportional term, integral term, and derivative term, respectively. If it is a curve, increase the gain of the proportional term of the PID controller to... Freeze the integral value or reduce the gain of the integral term to The PID output is obtained, and the path curvature is multiplied by the preset feedforward gain to generate an additional steering differential compensation amount, which is superimposed with the PID output as a motor control command to achieve advance and smooth steering control. If it is a crossroads, information collected by the vision module or color block sensor is used for identification to make a turning decision; If the path is lost, slowly rotate in place or backtrack along the last known direction and trajectory before the path was lost until a valid path signal is detected again.
[0014] The beneficial effects of this invention are: This invention achieves wide-range, high-density sampling of path edges using a dual-row, eight-channel grayscale sensor array, compared to traditional single-row or limited-sensor solutions. The symmetrical layout of the sensor array enables the system to effectively identify the lateral deviation direction and magnitude of the vehicle, maintaining stable recognition performance even under complex lighting conditions or when the road surface has slight stains or reflective interference. Combined with a PID closed-loop control algorithm, the system achieves millimeter-level path tracking accuracy and rapid dynamic response, ensuring smooth vehicle movement on straight lines, curves, and complex paths.
[0015] This invention achieves multi-sensor (vision, grayscale, color) fusion perception, low-cost embedded real-time control, and efficient collaboration between visual recognition and motion control. Attached Figure Description
[0016] Figure 1 This is a schematic diagram of the intelligent vehicle motion control system based on visual recognition and autonomous navigation according to an embodiment of the present invention.
[0017] Figure 2 This is a schematic diagram of the trajectory tracking method for an intelligent vehicle based on visual recognition and autonomous navigation according to an embodiment of the present invention. Detailed Implementation
[0018] The present invention will be further explained and described below with reference to the accompanying drawings and embodiments.
[0019] like Figure 1 As shown, this embodiment of the invention provides an intelligent vehicle motion control system based on visual recognition and autonomous navigation, including: a main control module and a vision module, a voice module, a remote control handle, a pathfinding module, a drive module, and a power module connected to the main control module.
[0020] The main control module uses an STM32F103C8T6 microcontroller, the vision module uses an OPENMV camera module, the voice module uses a DY-SV8F, the remote control uses a PS2 controller, the pathfinding module uses dual 4-channel grayscale and color block sensors, the drive module uses a WHEELTEC, and the power supply module uses a 12.6V lithium battery.
[0021] This intelligent vehicle motion control system possesses integrated intelligent operation capabilities, including autonomous pathfinding, visual recognition, voice interaction, and remote control. Each module communicates with the main control module through standardized interfaces. The system adopts a layered control architecture: the bottom layer is the hardware driver layer, responsible for data acquisition from various sensors and actuator control; the middle layer is the algorithm processing layer, containing core logic such as pathfinding algorithms and visual recognition algorithms; and the top layer is the application layer, implementing specific functions such as manual remote control, autonomous pathfinding, and voice interaction. The core hardware modules are shown in Table 1.
[0022] Table 1. Core Hardware Modules of the Smart Car The STM32F103C8T6 is based on the Arm Cortex-M3 core, with a maximum operating frequency of 72MHz, and features rich peripheral resources such as a 16-bit timer, SPI, I2C, and USART. Its core's single-cycle multiplication and hardware division capabilities provide computational support for real-time control algorithms; multiple serial port peripherals allow for parallel communication with voice modules, gamepads, driver modules, etc.; and the timer can precisely output PWM signals to control motor speed.
[0023] The pathfinding module includes eight grayscale sensors and color block sensors configured on the bottom of the smart car; the eight grayscale sensors are divided into two rows along the direction of travel, with four sensors evenly distributed laterally in each row; the color block sensors are located between the two rows of grayscale sensors or in front of the two rows of grayscale sensors.
[0024] This invention also provides a method for tracking the trajectory of an intelligent vehicle based on visual recognition and autonomous navigation. The algorithm primarily works by real-time monitoring of the relative positional relationship between the path and the vehicle, driving the motor to make dynamic adjustments to achieve accurate path tracking. The method includes the following steps: S101: Real-time acquisition of voltage values collected by 8 grayscale sensors and mapping them to grayscale values.
[0025] Eight grayscale sensors detect the intensity of reflected light from the road surface in real time, converting different colors (such as black tracks against a white background) into corresponding analog voltage signals, and then into digital signals. Assume each sensor outputs a voltage... Through normalization, the voltage values are mapped to normalized grayscale values. .
[0026] in: g i For the first i The grayscale sensor collects the grayscale value corresponding to the voltage. V i For the first i The voltage value collected by the grayscale sensor. V min and V max These are the voltage reference values corresponding to the black and white lines, respectively.
[0027] S102. Based on the position of the grayscale sensor, the normalized grayscale value of each row of grayscale sensors is weighted to obtain the weighted grayscale value of each row of grayscale sensors, and the lateral offset, yaw angle and path curvature of each row of grayscale sensors are calculated.
[0028] Let the lateral position of each sensor be... With the center of the sensor array as the origin, the position coordinates can be defined as: in: d The distance between adjacent grayscale sensors in each row. i =1...8 represents the grayscale sensor numbers, numbered from front to back. x i For the first i The lateral coordinates of the road grayscale sensor.
[0029] This invention employs a weighted average-based offset calculation, using Gaussian weights to obtain the weight coefficients for each grayscale sensor based on its position. in: To control the width of the weight distribution (usually taken as...) ); For the first i Weighting coefficients for road grayscale sensors.
[0030] The normalized gray values of each row of grayscale sensors are weighted based on weighting coefficients to obtain the weighted gray values of each row: Based on the weighted gray values, the lateral offset, yaw angle, and path curvature of each row are obtained; in: The weighted grayscale values for the front row. Weighted grayscale values for the front row; This represents the lateral deviation of the front-row grayscale sensors. The lateral deviation of the rear grayscale sensor G 0 represents the theoretical grayscale value of the path center; K e This is the offset gain coefficient, used to convert grayscale differences into physical offsets; θ Yaw angle; L This refers to the distance between the two rows of grayscale sensors. k For path curvature.
[0031] S103. Construct a state machine based on the gray values obtained from the 8 grayscale sensors to perform path feature recognition for the intelligent vehicle.
[0032] (1) The gray values of the obtained 8 gray sensors are binarized according to the threshold. If the value is higher than or equal to the threshold, it is set to 1, indicating that it is in the white area. If the value is lower than the threshold, it is set to 0, indicating that it is in the black area. The two rows of gray sensors obtain two sets of four-bit binary sequences, which form a spatial pattern vector.
[0033] (2) The state machine predefines five typical path features, including straight road state, starting to enter curve state, turning state, crossroads state and lost path state.
[0034] S_OnTrack (Straight Track State): Both rows of sensors stably detect a continuous path.
[0035] Entering Curve (S_EnteringCurve): The front sensor mode begins to change, and the mode vector shows asymmetrical characteristics, while the rear sensor is still in straight-line mode, indicating that the vehicle is entering a curve.
[0036] S_InCurve: Both rows of sensor modes indicate that the vehicle is in a curve, and the ef and er symbols are the same.
[0037] Crossing Status (S_Crossing): If the black signals in a row of sensors are discontinuous or distributed on both sides, it may indicate that an intersection has been encountered.
[0038] Lost path status S_Lost: Neither row of sensors detected a sufficiently strong path signal.
[0039] (3) Based on the spatial pattern vector obtained within the time period T, perform state machine judgment to obtain the path features of the current intelligent vehicle.
[0040] If both sets of four-bit binary sequences in the acquired spatial pattern vector are
[0011] or
[1100] , and remain unchanged within the time period T, then the current path characteristic of the intelligent vehicle is a straight road state.
[0041] If the second group of four-bit binary sequences in the acquired spatial pattern vector is still
[0011] or
[1100] , and the first group of four-bit binary sequences becomes
[0111] or
[1110] or
[0001] or
[1000] , and this continues for a period of time T, then the current path characteristic of the intelligent vehicle is that it is starting to enter a curve state.
[0042] If a set of four-bit binary sequences in the acquired spatial pattern vector shows discontinuous black / white signals or is distributed on both sides for a duration of T, then the current path characteristic of the intelligent vehicle is a crossroads state.
[0043] If the second group of four-bit binary sequences in the acquired spatial pattern vector are both
[0000] or
[1111] and last for a period of time T, then the current path feature of the intelligent vehicle is a lost path state.
[0044] S104. Based on the path characteristics, and using the lateral offset of each row of grayscale sensors, as well as the yaw angle and path curvature, output the control quantity.
[0045] This invention calculates lateral offset, yaw angle, and path curvature in real time. Unlike traditional PID control that only uses lateral offset, this improves overall path tracking performance, particularly cornering performance and stability. Typically, the forward offset *ef* is used as the primary control reference, representing the expected future position. That is, *e* = *ef*, which serves as the direct basis for the proportional, integral, and derivative terms, enabling basic feedback correction. The offset correction output from the PID controller is ultimately converted into a PWM duty cycle adjustment signal and applied to the motor drive module to control the speed difference between the left and right wheels, thereby achieving steering correction and path tracking for the vehicle.
[0046] If the road is straight, the lateral offset of the front grayscale sensor is converted into a motor control command by the PID controller; at this time, the yaw angle and path curvature are not considered.
[0047] Where u(t) is the control quantity, and Kp, Ki, and Kd are the control coefficients for the proportional, integral, and derivative terms, respectively.
[0048] u(t) is the control variable, where its sign represents the steering direction and its magnitude represents the correction strength. The proportional term directly generates the steering correction based on the current deviation; the integral term accumulates historical deviations to eliminate steady-state error; and the derivative term predicts the deviation change trend to suppress oscillations. The controller's output is ultimately converted into the speed difference between the left and right motors: when the vehicle veers to the right, the speed of the left wheel is increased or the speed of the right wheel is decreased, generating a steering torque to the left; and vice versa. Through continuous sensing, calculation, and control, the vehicle can dynamically adjust its driving direction to maintain stable travel along a predetermined path. The target speed v is converted into the corresponding PWM duty cycle D using a lookup table or linear mapping. Where Kv is the speed-duty cycle conversion coefficient, and D0 is the motor start-up threshold duty cycle. The STM32 generates two independent PWM signals through two timer channels TIM2_CH1 and TIM3_CH1, corresponding to the left and right wheels respectively.
[0049] If it is a curve, increase the proportional gain of the PID controller. To improve steering response speed, while freezing the integral value or reducing the gain of the integral term ( To prevent saturation, a PID output is obtained, and feedforward compensation is performed based on the real-time calculated path curvature. The path curvature is multiplied by a preset feedforward gain to generate an additional steering differential compensation amount, which is superimposed with the PID output as a motor control command to achieve advance and smooth steering control.
[0050] If it is a crossroads, information collected by the vision module or color block sensor is used for identification to make a turning decision; If the path is lost, a preset search will be initiated, using slow rotation in place or backtracking along the last known direction and trajectory before the path was lost, until a valid path signal is detected again.
[0051] Through this tightly coupled perception and decision-making mechanism, the state machine enables the system to go beyond simple reactive correction and acquire preliminary scene understanding and adaptive capabilities, thereby achieving higher robustness and smoother autonomous navigation performance in complex path environments.
[0052] Although embodiments of the present invention have been shown and described above, it is understood that the above embodiments are exemplary and should not be construed as limiting the present invention. Those skilled in the art can make changes, modifications, substitutions and variations to the above embodiments within the scope of the present invention.
Claims
1. A motion control system for an intelligent vehicle based on visual recognition and autonomous navigation, characterized in that, include: The main control module and the vision module, pathfinding module, drive module and power supply module connected to the main control module; The pathfinding module includes eight grayscale sensors and color block sensors configured on the bottom of the intelligent vehicle; the eight grayscale sensors are divided into two rows along the direction of travel, with four sensors evenly distributed laterally in each row; the color block sensors are located between the two rows of grayscale sensors or in front of the two rows of grayscale sensors. The main control module collects the outputs of the pathfinding module and the vision module in real time, and the generated control signals are connected to the drive module to drive the motor of the intelligent car to run and control the driving speed and direction. The power module supplies power to the intelligent vehicle's motion control system.
2. The intelligent vehicle motion control system based on visual recognition and autonomous navigation according to claim 1, characterized in that: It also includes a voice module and a remote control handle connected to the main control module; The main control module receives commands from the remote control handle and status feedback from the voice module to control the movement of the smart car.
3. The intelligent vehicle motion control system based on visual recognition and autonomous navigation according to claim 1, characterized in that: The main control module uses STM32F103C8T6, and the vision module uses OpenMV.
4. A method for tracking the trajectory of an intelligent vehicle based on visual recognition and autonomous navigation, based on the intelligent vehicle motion control system based on visual recognition and autonomous navigation as described in claims 1-3, characterized in that, Includes the following steps: (1) Real-time acquisition of voltage values collected by 8 grayscale sensors and mapping them to normalized grayscale values; (2) Based on the position of the grayscale sensor, the normalized grayscale value of each row of grayscale sensors is weighted to obtain the weighted grayscale value of each row of grayscale sensors, and the lateral offset, yaw angle and path curvature of each row of grayscale sensors are calculated. (3) Construct a state machine based on the gray values obtained from the 8 grayscale sensors to identify the path features of the intelligent vehicle; (4) Based on the path characteristics, output control quantity based on the lateral offset of each row of grayscale sensors, as well as the yaw angle and path curvature.
5. The intelligent vehicle trajectory tracking method based on visual recognition and autonomous navigation according to claim 1, characterized in that, The specific steps (1) are as follows: in: g i For the first i The grayscale sensor collects the normalized grayscale value corresponding to the voltage. V i For the first i The voltage value collected by the grayscale sensor. V min and V max These are the voltage reference values corresponding to the black and white lines, respectively.
6. The intelligent vehicle trajectory tracking method based on visual recognition and autonomous navigation according to claim 1, characterized in that, Step (2) specifically involves: Eight grayscale sensors are arranged in two rows along the travel path, with four sensors evenly distributed laterally in each row. Taking the center of the grayscale sensor array as the origin, the lateral coordinates of the eight grayscale sensors are: in: d The distance between adjacent grayscale sensors in each row. i =1...8, which are the numbers of the grayscale sensors, numbered from the front row to the back row; x i For the first i The lateral coordinates of the road grayscale sensor; Based on the location of each grayscale sensor, a Gaussian weight is applied to obtain the weight coefficient for each grayscale sensor: in: To control the width of the weight distribution; For the first i Weighting coefficients for road grayscale sensors; The normalized gray values of each row of grayscale sensors are weighted based on weighting coefficients to obtain the weighted gray values of each row: Based on the weighted gray values, the lateral offset, yaw angle, and path curvature of each row are obtained; in: The weighted grayscale values for the front row. The weighted grayscale values for the front row; g i For the first i The grayscale sensor collects the normalized grayscale value corresponding to the voltage. This represents the lateral deviation of the front-row grayscale sensors. The lateral deviation of the rear grayscale sensor G 0 represents the theoretical grayscale value of the path center; K e This is the offset gain coefficient, used to convert grayscale differences into physical offsets; θ Yaw angle; L This refers to the distance between the two rows of grayscale sensors. K For path curvature.
7. The intelligent vehicle trajectory tracking method based on visual recognition and autonomous navigation according to claim 1, characterized in that, Step (3) specifically involves: (3.1) The gray values of the obtained 8 gray sensors are binarized according to the threshold. If the gray value is higher than or equal to the threshold, it is set to 1, indicating that it is in the white area. If it is lower than the threshold, it is set to 0, indicating that it is in the black area. The two rows of gray sensors obtain two sets of four-bit binary sequences, which form a spatial pattern vector. (3.2) The state machine predefines five typical path characteristics, including straight road state, starting to enter a curve state, turning state, crossroads state, and lost path state: (3.3) Based on the spatial pattern vector obtained within the time period T, perform state machine judgment to obtain the path features of the current intelligent vehicle.
8. The intelligent vehicle trajectory tracking method based on visual recognition and autonomous navigation according to claim 7, characterized in that: If both sets of four-bit binary sequences in the acquired spatial pattern vector are [0011] or [1100], and remain unchanged within the time period T, then the current path feature of the intelligent vehicle is a straight road state. If the second group of four-bit binary sequence in the acquired spatial pattern vector is still [0011] or [1100], and the first group of four-bit binary sequence becomes [0111] or [1110] or [0001] or [1000], and this lasts for a period of time T, then the current path characteristic of the intelligent vehicle is that it is starting to enter a curve state. If a set of four-bit binary sequences in the acquired spatial pattern vector shows discontinuous black / white signals or is distributed on both sides for a duration of T, then the current path characteristic of the intelligent vehicle is a crossroads state. If the second group of four-bit binary sequences in the acquired spatial pattern vector are both [0000] or [1111] and last for a period of time T, then the current path feature of the intelligent vehicle is a lost path state.
9. The intelligent vehicle trajectory tracking method based on visual recognition and autonomous navigation according to claim 8, characterized in that: If the road is straight, the lateral offset of the front grayscale sensor is converted into a motor control command by the PID controller; at this time, the yaw angle and path curvature are not considered. in, u (t) is the control variable. K p , K i , K d These are the gains of the proportional term, integral term, and derivative term, respectively. If it is a curve, increase the gain of the proportional term of the PID controller to... Freeze the integral value or reduce the gain of the integral term to The PID output is obtained, and the path curvature is multiplied by the preset feedforward gain to generate an additional steering differential compensation amount, which is superimposed with the PID output as a motor control command to achieve advance and smooth steering control. If it is a crossroads, information collected by the vision module or color block sensor is used for identification to make a turning decision; If the path is lost, slowly rotate in place or backtrack along the last known direction and trajectory before the path was lost until a valid path signal is detected again.