An intelligent road marking robot fusing AI recognition and embedded control

Abstract

The application discloses a kind of intelligent road marking machine of fusion AI identification and embedded control, comprising: vehicle body walking system, ground cleaning module, gantry type spraying execution mechanism, AI sensing early warning system, embedded control system, bluetooth man-machine interaction module and power management module;Vehicle body walking system adopts four-wheel differential drive structure;Ground cleaning module sweeps road dust, chippings, and cleaning action is linked with vehicle body walking start-stop;Gantry type spraying execution mechanism drives spray head to move in two-dimensional plane;AI sensing early warning system includes edge computing platform, depth camera, audible and visual alarm unit;Embedded control system adopts STM32 microcontroller and edge computing platform to constitute double main control architecture.The application can realize personnel real-time detection early warning, ground litter cleaning, gantry double-shaft accurate spraying, embedded motion control, bluetooth remote interaction integrated operation, significantly improve construction efficiency, safety, marking precision and marking quality.

Description

Technical Field

[0001] This invention belongs to the field of road construction equipment and intelligent control technology, specifically involving an intelligent road marking robot that integrates AI recognition and embedded control, combining machine vision, deep learning recognition, embedded control and automatic spraying technology. Background Technology

[0002] Road markings are crucial facilities for road traffic management and safety guidance. Currently, road marking construction still relies primarily on manual labor using hand-pushed carts, which is inefficient, labor-intensive, and the quality depends heavily on manual experience. Furthermore, operations at night, in high temperatures, and in densely trafficked areas pose significant safety risks. Existing marking equipment has a low level of automation, relying heavily on manual visual control of the trajectory, making it prone to deviation. A few automatic marking devices only have basic walking functions and lack personnel safety detection and proactive warning capabilities. Additionally, the lack of ground cleaning mechanisms before spraying results in poor adhesion and easy detachment of the markings.

[0003] With the rapid expansion of urban transportation networks and the improvement of infrastructure construction, road markings, as an important component of traffic organization and safety management, are receiving increasing attention for their construction efficiency and quality. Currently, most road markings still rely on manual operation, which is not only inefficient and labor-intensive but also poses significant safety risks and inconsistent accuracy, failing to meet the demands of modern roads for intelligent and standardized construction. Especially during nighttime construction, high-temperature operations, or in areas with heavy traffic, manual marking work faces significant safety hazards, urgently requiring the introduction of intelligent equipment to assist or replace traditional methods. For example... Figure 7 The image shows the current state of manual line marking construction.

[0004] In recent years, the rapid development of artificial intelligence, machine vision, and embedded control technologies has provided crucial support for the intelligent transformation of road construction. Addressing the current situation where road marking work is largely done manually using hand-pushed carts and the automation level of these carts is low, Yang Cheng et al. proposed a design scheme for an automated road marking cart system based on microcontroller control. [1] Furthermore, current road marking vehicles primarily rely on human visual observation to operate and spray markings along a baseline. However, driver fatigue can easily lead to the vehicle veering off course. Yuan Weiqi et al. proposed a machine vision-based method for detecting road marking vehicle tracking of the baseline and developed a corresponding baseline detection device that can provide alerts when the vehicle deviates from its course. [2]Meanwhile, with the rapid development of highway construction in my country in recent years, the requirements for road marking equipment have also increased. However, the automation level of domestic road marking equipment is relatively low, and the accuracy and efficiency of marking largely depend on the experience of construction personnel. Therefore, based on the working characteristics of road marking machines, Li Hong et al. proposed an adaptive threshold method to address the difficulties of incomplete road boundary detection and inaccurate obstacle detection. This method provides theoretical support for the detection and tracking of road boundaries by road marking machines. [3] Furthermore, research has been conducted on the automation of hot-melt line marking machines during new road marking construction. To improve marking accuracy and automation levels, existing research has adopted a combination of STM32 main control and visual navigation sensors to establish an automated execution system and path tracking control system, achieving precise operation of the line marking machine during new road marking construction. [4] Then, to improve the path accuracy and trajectory smoothness of road marking robots, existing research has adopted a combined navigation system integrating RTK and inertial navigation, and achieved fine control of lateral deviation and trajectory smoothness through error state extended Kalman filtering and model predictive control methods. [5] Currently, road marking operations in my country are mainly carried out manually, which suffers from high labor intensity, low work efficiency, and the quality of road markings depends heavily on the skills of the construction workers. Therefore, it is essential to conduct in-depth research on intelligent and unmanned road marking equipment. [6]

[0005] In recent years, embedded platforms, depth cameras, and deep learning models have provided the technological foundation for small-scale intelligent construction equipment. However, a fully integrated intelligent road marking system that combines AI-powered personnel recognition and early warning, ground cleaning, high-precision dual-axis spraying, embedded motion control, and remote interaction has yet to be developed. Therefore, there is an urgent need for an automated, safe, high-precision, and remotely controllable intelligent road marking machine to address the technical pain points of existing technologies, such as low efficiency, poor safety, insufficient accuracy, and unsatisfactory marking quality. Summary of the Invention

[0006] The purpose of this invention is to provide an intelligent road marking robot that integrates AI recognition and embedded control, enabling real-time personnel detection and early warning, ground debris cleaning, gantry dual-axis precision spraying, embedded motion control, and Bluetooth remote interaction, significantly improving construction efficiency, safety, marking accuracy, and marking quality, thereby solving the problems mentioned in the background art.

[0007] To achieve the above objectives, the present invention provides the following technical solution: an intelligent road marking robot integrating AI recognition and embedded control, comprising: a vehicle walking system, a ground cleaning module, a gantry spraying execution mechanism, an AI perception and early warning system, an embedded control system, a Bluetooth human-machine interaction module, and a power management module;

[0008] The vehicle's walking system adopts a four-wheel differential drive structure, including a frame, drive wheels and follower wheels mounted on the frame, and a DC geared motor that drives the drive wheels;

[0009] The ground cleaning module is located at the front of the vehicle frame and includes a cleaning frame, a detachable roller brush mounted on the cleaning frame, and multiple DC cleaning motors that drive the detachable roller brush. The multiple DC cleaning motors drive the detachable roller brush to rotate and clean the road surface dust and debris. The cleaning action is linked to the vehicle's walking system for starting and stopping.

[0010] The gantry-type spraying actuator is installed in the middle of the vehicle frame and includes a gantry frame, an atomizing nozzle mounted on the gantry frame, a micro water pump, a water tank, and a leak-proof device. The gantry frame includes an X-axis slide rail, a Y-axis slide rail, an X-axis stepper motor, a Y-axis stepper motor, and a ball screw. The X-axis slide rail is arranged laterally, and the Y-axis slide rail is slidably connected to the X-axis slide rail. The atomizing nozzle is slidably connected to the Y-axis slide rail. The X-axis stepper motor and the Y-axis stepper motor drive the atomizing nozzle to move in a two-dimensional plane through the ball screw. The micro water pump draws pigment from the water tank and sprays it through the atomizing nozzle. The leak-proof device enables immediate shutdown.

[0011] The AI ​​perception and early warning system is installed on the upper part of the vehicle frame and includes a JetsonNano edge computing platform, an AstraS depth camera, and an audible and visual alarm unit. The AstraS depth camera collects color images and depth information of the construction area in real time. The JetsonNano deploys a lightweight YOLOv5 target detection model to identify personnel targets in real time. When personnel enter the preset safe area, the audible and visual alarm unit is triggered, and the early warning response delay is less than or equal to 500ms.

[0012] The embedded control system uses an STM32 microcontroller as its underlying core, forming a dual-master architecture with a Jetson Nano. The STM32 is responsible for motion control and logic scheduling for walking, sweeping, and spraying; the Jetson Nano is responsible for image recognition and early warning decisions. The two communicate via serial port and use CRC check to ensure data reliability.

[0013] Preferably, the frame is made of aluminum alloy profile, and the detachable roller brush is made of flexible nylon; the DC geared motor is controlled by an embedded control system to realize forward, reverse, steering and constant speed cruise, with a travel speed of 0.2m / s.

[0014] Preferably, the atomizing nozzle can spray markings with a width of 5-8cm, a spraying movement speed of 0.05m / s, and supports spraying of straight lines, diagonal lines, and herringbone patterns, with a spraying trajectory error of less than or equal to ±1mm.

[0015] Preferably, the AstraS depth camera has an effective recognition distance of 0.5 to 3 meters and a personnel recognition accuracy of no less than 95%.

[0016] Preferably, the robot has overall dimensions of 100cm×72cm×67cm and a weight of less than or equal to 20kg.

[0017] Preferably, the Bluetooth human-machine interaction module adopts an HC-05 Bluetooth module, which is connected to the STM32 to receive mobile APP commands and realize remote start / stop, trajectory setting, spraying control and status feedback.

[0018] Preferably, the power management module adopts dual independent power supply: the 5V lithium battery pack powers the Jetson Nano, and the 12V lithium battery pack powers the STM32, motor, water pump and alarm unit, with a continuous working battery life of not less than 1 hour.

[0019] Preferably, the machine uses a differential-driven MG513 motor as the main walking mechanism, the total mass of the machine is M=20kg, the rolling friction coefficient is μ=0.02, the wheel radius is r=0.09m, and the expected maximum speed is v=0.2m / s;

[0020] According to the dynamic formula:

[0021]

[0022] Because the driving force needs to be greater than the rolling friction, to ensure a starting margin, the required driving force is calculated as 1.5 times:

[0023]

[0024] The required motor output torque is:

[0025]

[0026] The MG531 motor output shaft reduction ratio is i=30:1 for torque increase. Therefore, the rated torque of the DC geared motor should be between 0.3 and 0.5 N·m to meet the load requirements. For maximum speed calculation:

[0027]

[0028]

[0029] The motor output shaft should provide a speed of about 20 rpm after deceleration, while the MG531 speed is 293 rpm. According to calculations, the MG531 geared DC motor (12V 30:1) can meet the usage requirements.

[0030] Preferably, the gantry-type spraying actuator uses a transverse load-bearing beam, i.e., the X-axis guide rail, on which the spraying slider and the Y-axis linear slide are mounted. The beam is made of aluminum alloy profile, and its mechanical performance parameters are as follows:

[0031] elastic modulus

[0032] Yield strength

[0033] Safety factor

[0034] Effective span of beam

[0035] rectangular aluminum profile

[0036] The total mass of the spraying module, guide rail slider, spray head, etc. is approximately All loads are concentrated at the midpoint of the beam, and it is considered as a simply supported beam with a concentrated load at the midpoint. Gravity load:

[0037]

[0038] The maximum bending moment occurs at the midpoint of the beam:

[0039]

[0040] Moment of inertia of rectangular section:

[0041]

[0042] Maximum deflection (concentrated force at the midpoint of a simply supported beam):

[0043]

[0044] Within the structural tolerance range of ±1mm, it meets the usage requirements;

[0045] The maximum normal stress in the cross section occurs at the lower edge of the middle part of the beam:

[0046]

[0047] in,

[0048]

[0049] The yield strength is 276 MPa, taking into account the safety factor. The allowable stress is:

[0050]

[0051] Obviously:

[0052]

[0053] The safety margin is extremely high, and the beam fully meets the strength requirements.

[0054] Preferably, the gantry-type spraying actuator uses a micro water pump to draw water from the water tank and atomize it through the nozzle to simulate road marking operations. The required water pump flow rate is [not specified]. The nozzle outlet diameter is ;

[0055] The theoretical outlet velocity of the nozzle (continuous flow, Bernoulli estimate) is:

[0056]

[0057] To achieve this speed, the minimum head required, ignoring energy loss, is:

[0058]

[0059] Considering system pressure loss, select the appropriate head. DC water pump.

[0060] Compared with the prior art, the beneficial effects of the present invention are:

[0061] 1. AI Safety Early Warning: Integrating depth cameras and YOLOv5 models, it can identify construction workers in real time and provide sound and light warnings, filling the safety protection gap of existing line marking equipment and reducing construction risks;

[0062] 2. High-precision spraying: Gantry dual-axis linkage + stepper motor ball screw drive, high trajectory accuracy, supports complex patterns, significantly better than manual scribing;

[0063] 3. Improved pavement marking quality: The front roller brush removes debris, and the anti-leakage device prevents dripping, enhancing the adhesion and uniformity of the pavement markings;

[0064] 4. Remote and convenient control: Bluetooth mobile APP control, simple operation, adaptable to various construction scenarios;

[0065] 5. Compact structure and controllable cost: The lightweight aluminum alloy design is small in size, light in weight, easy to deploy, and has good engineering implementation. Attached Figure Description

[0066] Figure 1This is a schematic diagram of the overall structure of the present invention;

[0067] Figure 2 This is a front view of the present invention.

[0068] Figure 3 This is a side view of the present invention;

[0069] Figure 4 This is a functional breakdown diagram of the present invention;

[0070] Figure 5 This is a schematic diagram of the gantry frame structure in this invention;

[0071] Figure 6 This is a schematic diagram of the anti-leakage device in this invention;

[0072] Figure 7 This refers to manually drawn construction drawings in the background art of this invention;

[0073] In the diagram: 1. DC geared motor; 2. Drive wheel; 3. Warning light; 4. Water tank; 5. First sweeping motor; 6. Gantry frame; 7. Second sweeping motor; 8. Swing bracket; 9. Sweeping bracket; 10. Fixed bracket; 11. Frame; 12. Miniature water pump; 13. Detachable roller brush; 14. Floor sweeping module; 15. X-axis slide rail; 16. Y-axis slide rail; 17. X-axis stepper motor; 18. Y-axis stepper motor; 19. Coupling; 20. Nozzle fixing component; 21. Follower wheel; 22. Ball screw; 23. Valve seat; 24. Sealing ball; 25. Spring; 26. Liquid inlet head; 27. Baffle ball. Detailed Implementation

[0074] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0075] Example 1

[0076] like Figures 1-3 As shown, an intelligent road marking robot integrating AI recognition and embedded control includes: a vehicle walking system, a ground cleaning module 14, a gantry spraying actuator, an AI perception and early warning system, an embedded control system, a Bluetooth human-machine interaction module, and a power management module.

[0077] The vehicle's running system adopts a four-wheel differential drive structure, including a frame, drive wheels 2 and follower wheels 21 mounted on the frame, and a DC geared motor 1 that drives the drive wheels 2;

[0078] The ground cleaning module 14 is located at the front of the vehicle frame and includes a cleaning frame, a detachable roller brush 13 mounted on the cleaning frame, and a first cleaning motor 5 and a second cleaning motor 7 that drive the detachable roller brush 13. The cleaning frame includes a fixed bracket 10 fixed to the vehicle frame, a swing bracket 8 connected to the fixed bracket 10, and a cleaning bracket 9 connected to the swing bracket 8. The first cleaning motor 5 drives the swing bracket 8 to swing back and forth through a gear set, so that the detachable roller brush 13 forms a fan-shaped sweeping area in front of the vehicle frame 11. The second cleaning motor 7 drives the cleaning bracket 9 to tilt and rotate through a gear set to adjust the contact angle between the detachable roller brush 13 and the surface to be cleaned. The detachable roller brush 13 itself does not rotate actively, but generates sweeping motion with the swing of the swing bracket and the angle adjustment of the cleaning bracket, thereby cleaning dust and debris on the road surface. The cleaning action is linked to the start and stop of the vehicle's walking system.

[0079] The intelligent road marking robot of this invention aims to achieve functions such as intelligent perception, safety warning, path marking, automatic spraying and cleaning of construction areas (e.g., Figure 4 As shown in the figure, it is applicable to scenarios such as road construction and factory planning. The system structure adopts a modular design concept and mainly consists of five core systems: perception system, execution system (including spraying and cleaning), walking system, control system, and human-machine interaction system.

[0080] like Figure 5 As shown, the gantry-type spraying actuator is installed in the middle of the vehicle frame, including a gantry frame 6, an atomizing nozzle mounted on the gantry frame 6, a micro water pump 12, a water tank 4, and a leak-proof device. The gantry frame 6 includes an X-axis slide rail 15, a Y-axis slide rail 16, an X-axis stepper motor 17, a Y-axis stepper motor 18, a coupling 19, and a ball screw 22. The X-axis slide rail 15 is arranged laterally, and the Y-axis slide rail 16 is slidably connected to the X-axis slide rail 15. The atomizing nozzle is slidably connected to the Y-axis slide rail 16. The X-axis stepper motor 17 and the Y-axis stepper motor 18 drive the atomizing nozzle to move in a two-dimensional plane through the ball screw. The micro water pump 12 draws pigment from the water tank 4 and sprays it through the atomizing nozzle. The leak-proof device enables immediate shutdown.

[0081] The system uses Jetson Nano as the vision control platform, which is responsible for running AI algorithms, detecting personnel in the construction area and triggering alarms; the STM32 microcontroller serves as the motion control core, completing the movement of the vehicle and the spraying control; the spraying mechanism adopts a gantry-type dual-axis structure to achieve high-precision path spraying; the front cleaning mechanism is used to clean the path; the human-machine interaction uses mobile phone Bluetooth control to achieve convenient scheduling and operation.

[0082] Meanwhile, the entire machine's mechanism design was completed using 3D modeling software (such as SolidWorks). The overall body uses aluminum alloy profiles to construct the frame, ensuring lightweight and structural stability. The upper part integrates the sensing and processing module, the middle part is the gantry spraying mechanism, the front part is the roller brush cleaning system, and the lower part is the drive wheel, with a compact layout and reasonable center of gravity. The overall size of the machine is controlled within 110cm × 72cm × 67cm, which facilitates transportation and deployment.

[0083] The perception system uses Jetson Nano as an embedded AI computing platform, paired with an AstraS depth camera to acquire images of the surrounding environment. A YOLOv5 deep learning model is deployed to identify the presence of personnel in the images in real time. When personnel are detected approaching the work area, the system activates an alarm via alarm light 3 and a voice module, alerting personnel to take evasive action and improving construction safety. The AstraS depth camera provides color images and depth information, offering high recognition accuracy within 3 meters. Jetson Nano running the YOLOv5 model exhibits excellent real-time performance, with recognition latency controlled within 500ms.

[0084] The gantry-type spraying actuator is a key component for intelligent road marking robots to complete their core task—line marking—and directly affects the quality and efficiency of the construction results. This invention adopts a gantry-type two-axis linkage structure design for the spraying system, combining a stepper motor, slide rail, nozzle, and water pump module to form a stable, efficient, and scalable road marking actuator.

[0085] In terms of mechanical structure, the gantry-type spraying actuator is mounted on top of the robot body. Through a gantry structure composed of two-stage slide rails along the X and Y axes, it enables flexible movement of the spray head within a two-dimensional plane. This structure features high positioning accuracy and strong structural stability, making it particularly suitable for spraying regular patterns on flat surfaces. The drive unit uses a high-precision stepper motor in conjunction with a ball screw drive to achieve precise control of the spraying trajectory. The displacement range of each axis is set according to the overall machine width and task requirements, generally supporting a 30cm × 30cm two-dimensional movement area, capable of covering common construction patterns such as straight lines and herringbone markings.

[0086] Regarding the spraying medium, to reduce costs and environmental impact, this invention uses acrylic paint instead of traditional road marking paint for simulated spraying. The nozzle is connected to a miniature water pump, which draws acrylic paint from a water tank and discharges it through a pre-designed leak-proof device (such as...). Figure 6As shown in the diagram, precise spraying is achieved by coordinating with the two-axis motion trajectory. The leak-proof device is vertically designed on the water pipe between the pump and the tank, including a valve seat 23, a sealing ball 24 inside the valve seat 23, a spring 25, and a limiting seat 26 and a stop ball 27 located above the valve seat 23. In the non-working state (pump not working): the spring presses the sealing ball → the ball adheres to the valve seat → the passage is closed → preventing leakage; in the working state (pump working): liquid pressure pushes open the sealing ball → the ball leaves the valve seat → the passage opens → liquid flows out; when the pump stops supplying liquid (pump stops working): the pressure drops → the spring pushes the sealing ball back to its original position → the valve seat re-seals → preventing residual liquid dripping; this allows for immediate shut-off and shutdown, avoiding liquid dripping that could affect the final spraying effect. The function of the stop ball 27 is that when the pump is working, the liquid flows upwards through the limiting seat, causing the stop ball to be pushed open and allowing liquid to pass through. When the pump stops working, the stop ball, under its own weight, falls downwards and blocks the limiting seat, preventing further liquid leakage, thus achieving secondary leak prevention.

[0087] In practice, to improve the uniformity and adhesion of the spray, a fine atomizing spray nozzle was selected. The atomization angle and flow rate were adjusted through multiple rounds of experiments to match the robot's travel speed and the required line width. The line width was controlled between 5 and 8 cm to meet standard construction specifications, while also allowing for replacement and adjustment as needed.

[0088] The entire spraying module supports both automatic and manual working modes: in automatic mode, the spraying trajectory is controlled by a preset program, which is suitable for standardized construction tasks; in manual mode, it is remotely controlled via Bluetooth, and the operator can manually trigger the spraying action through a mobile APP for on-site debugging or emergency operation.

[0089] To ensure the quality and adhesion of road markings, this invention incorporates a dedicated ground cleaning module at the front end of the intelligent road marking robot. The main task of this module is to remove dust, debris, fallen leaves, and other obstructions from the road surface before spraying, ensuring a clean and smooth surface for the nozzle, thereby improving the clarity and durability of the markings. This module not only optimizes the construction effect but also enhances the robot's adaptability to complex construction environments, making it a crucial component of the intelligent road marking process.

[0090] This module employs a mechanical roller brush structure for floor cleaning. The roller brush is mounted at the front of the vehicle, close to the ground, with its rotation axis perpendicular to the robot's direction of travel. The roller brush is made of flexible nylon bristles, balancing cleaning efficiency with floor protection; it effectively removes lightweight debris without damaging the floor coating or structure. During cleaning, the brush roller is driven by an independent DC motor, rotating at a constant speed in coordination with the robot's direction of travel to complete continuous cleaning operations.

[0091] In terms of drive and control, the sweeping motor is controlled by an STM32 main control board, supporting constant speed operation and start-up delay logic. To avoid energy waste and unnecessary wear caused by the brush rollers idling, a linkage mechanism with the vehicle's movement status is incorporated into the system design: when the robot starts moving forward, the sweeping motor automatically starts; when the robot stops or performs on-site debugging operations, the brush rollers stop running. This linkage mechanism is implemented through the main control system's logic control of the motor drive board, avoiding complex sensor triggering designs and improving system stability and response efficiency. The module has a compact structure, is lightweight, and facilitates vehicle assembly and maintenance. The sweeping device adopts a modular design, allowing for quick disassembly, replacement, or upgrades according to actual usage scenarios. If future expansion to more complex construction site applications is possible, optional dust collection components, negative pressure filtration systems, etc., can be added to further improve the ground treatment effect. It demonstrates good operational adaptability and continuous stability in various typical scenarios (such as areas covered with fallen leaves and areas with residual construction dust). In addition, the introduction of the ground sweeping module also enhances the overall intelligence and automation level of the robot, making its application in intelligent construction scenarios more complete and professional.

[0092] The vehicle's walking system is the foundation of the robot's movement and navigation. Its structural design and drive method directly affect the robot's mobility, stability, and construction accuracy within the construction area. In designing the walking mechanism, we focused on the robot's adaptability to complex outdoor road conditions, the balance of power output, and its compatibility with the control system. Ultimately, we selected a four-wheel structure based on differential drive mode, paired with a DC motor, to achieve flexible and efficient movement control.

[0093] In terms of specific structure, the vehicle adopts a symmetrical four-wheel layout, with the front and rear wheels supported by aluminum profiles and rubber tires providing good grip and shock absorption. The drive system uses front-wheel differential drive, meaning each front wheel is independently driven by a DC geared motor. Turning or rotation on the spot is achieved by adjusting the speed difference between the two motors. This design eliminates the need for a servo or steering mechanism, resulting in a simple mechanical structure and rapid response, making it particularly suitable for construction scenarios with limited space or frequent direction adjustments. The rear wheels are free-rotating follower wheels, adapting to different travel directions and effectively reducing the turning radius.

[0094] To ensure the robot's stability in various terrains, this invention employs a high-torque, low-speed metal geared DC motor, providing ample output torque suitable for low-speed, heavy-load operation. Simultaneously, the chassis utilizes a lightweight aluminum alloy frame, and the connection points between the motor and wheel system are reinforced to enhance the overall structural strength.

[0095] In terms of control strategy, the walking mechanism uses an STM32 controller to control the H-bridge motor drive board via PWM signals to implement motion commands such as forward, backward, and turning. The main control program is designed with multiple control modes, including manual control, fixed-path control, and obstacle avoidance mode. In manual mode, the user sends motion commands via Bluetooth on their mobile phone, and the robot can respond in real time with actions such as forward movement, turning, and stopping. In fixed-path control, the robot completes marking work according to a preset route, operating synchronously with the painting system to achieve automated operation. Furthermore, the system reserves expansion interfaces for installing LiDAR or ultrasonic modules, allowing for the future introduction of SLAM positioning or intelligent obstacle avoidance algorithms to further enhance autonomous driving capabilities.

[0096] The vehicle's powertrain and power management have also been rationally configured. The motor drive is powered by an independent 12V lithium battery pack, ensuring sufficient range and operational stability. During extended operation, the system supports external charging modules or battery replacement to ensure uninterrupted mission operation.

[0097] Experimental tests show that the walking mechanism can achieve a maximum travel speed of approximately 0.4 meters per second, possessing stable straight-line travel capability and precise on-the-spot turning performance. It demonstrates good adaptability on standard concrete roads, asphalt roads, and slightly undulating terrain. Thanks to the differential speed control mechanism, the robot can flexibly respond to various changes in work paths, meeting the needs of complex construction routes such as straight lines, curves, and point-to-point turning.

[0098] The control system is the "brain" of the intelligent road marking robot, enabling coordinated operation and intelligent decision-making across multiple modules. It undertakes key functions such as instruction parsing, status monitoring, path planning, and task execution. To meet the system's high requirements for response speed, execution accuracy, and module compatibility, a "dual-master controller + multi-submodule collaborative" control architecture is adopted. The STM32 microcontroller serves as the underlying real-time control core, while the Jetson Nano edge computing platform handles high-performance image recognition and intelligent sensing. The two operate in coordination via serial communication, forming a deeply integrated hardware and software intelligent control system.

[0099] The STM32 control module is primarily responsible for the execution and scheduling of low-level tasks, including motor control, spraying actions, ground cleaning devices, and the coordinated management of the voice / light alarm system. Its programming is done in bare-metal C language to improve execution efficiency and response speed. Through PWM speed regulation and GPIO output control, the STM32 can precisely control the differential motor speed, enabling stable robot movement and flexible turning. Simultaneously, it meticulously schedules the stepper motors of the spraying system to ensure synchronized nozzle movements and accurate spraying trajectories. Furthermore, the STM32 is also responsible for receiving and parsing data from the Bluetooth module, allowing users to issue movement or task commands via a mobile app, enhancing the convenience of human-machine interaction.

[0100] JetsonNano, serving as a high-level perception and intelligent decision-making platform, integrates a deep learning inference environment and a camera data processing module. It deploys a YOLOv5 object detection model, combined with an AstraS depth camera, to identify the presence of personnel within the construction area. Once personnel are detected entering the preset safety boundary, JetsonNano immediately triggers an early warning mechanism and sends an interrupt signal to the STM32 microcontroller via serial port. The STM32 then controls a buzzer, warning light, or voice broadcast module to issue a warning. This mechanism effectively ensures the safety of personnel at the construction site, demonstrating the robot's "intelligent early warning" capability.

[0101] In control systems, data communication mechanisms are particularly crucial. The Jetson Nano and STM32 communicate via serial port (UART) for command exchange, employing CRC checksums to ensure data reliability and prevent crosstalk or false triggering. Internally, the STM32 uses a state machine to manage different task flows, such as the logical sequence of "standby - start cleaning - paint preparation - execute paint - end reset," ensuring the orderly execution of tasks in stages. Sensor signals (such as servo position feedback, brush speed, and water spray status) are acquired through the ADC interface and the system's operating status is fed back to the user via Bluetooth.

[0102] In terms of power management, the control system is powered by two independent power supplies: the Jetson Nano uses a high-capacity lithium battery pack to provide a stable 5V power supply; the STM32 and its control peripherals (such as motor drives, servos, solenoid valves, etc.) are powered by independent 12V battery modules, with a step-down module to convert them to the required voltage level. This distributed power supply strategy avoids power interference between modules and improves system stability and maintainability.

[0103] In terms of functional expansion, the control system reserves multiple I / O interfaces and I2C and SPI communication ports, facilitating the future integration of peripherals such as LiDAR, ultrasonic ranging, and GPS positioning, and gradually upgrading the robot from manual control to fully automated path planning and environmental perception navigation. Furthermore, by building a simple WebSocket communication interface, future functions such as wireless remote control over a local area network, task uploading, and status monitoring can be realized, further expanding its engineering practicality.

[0104] After the intelligent road marking robot of this invention is activated as a construction platform, it continuously collects environmental image data of the construction area using an onboard depth camera, and processes the collected image data using a deep learning target detection algorithm to identify personnel targets within the construction area in real time. The identification results include the category of the personnel target and information on its location, which are used to determine the safety status of the work area.

[0105] During the line marking process, the robot uses an embedded control system to drive the walking mechanism to move along a preset path and assigns time stamp information to the motion data during the movement to ensure the consistency of the walking status, spraying action and recognition results in the time dimension.

[0106] The acquired environmental image data is processed for abnormal personnel detection to identify personnel approaching the work area and obtain their category information and location within the image. When the same person is detected in multiple consecutive frames, the detection results can be merged based on target category consistency and spatial distance thresholds to form a stable safety warning event.

[0107] Based on the timestamp information of the image data, safety warning events are associated with the robot's operating status information at the corresponding time to determine the danger warning level and trigger an audible and visual alarm. The warning information can be represented in at least one of the following forms: danger distance, danger area level, or work stoppage instruction.

[0108] Finally, the embedded control system drives the gantry spraying mechanism to spray the markings along a preset trajectory, while simultaneously activating the front-end sweeping device to remove debris from the road surface. After the marking operation is completed, the spraying trajectory, identification records, and operational status data are automatically saved, generating a traceable construction record and anomaly warning results. Construction personnel can quickly review the marking quality and safety incidents based on these results, and reconfirm or supplement the work area.

[0109] Example 2

[0110] This invention drives the system design

[0111] The machine uses a differential-driven MG513 motor as the main walking mechanism. The total mass of the machine is M=20kg, the rolling friction coefficient is μ=0.02, the wheel radius is r=0.09m, and the expected maximum speed is v=0.2m / s.

[0112] According to the dynamic formula:

[0113]

[0114] Because the driving force needs to be greater than the rolling friction, to ensure a starting margin, the required driving force is calculated as 1.5 times:

[0115]

[0116] The required motor output torque is:

[0117]

[0118] The MG531 motor output shaft reduction ratio is i=30:1 for torque amplification. Therefore, the rated torque of the DC geared motor should be between 0.3 and 0.5 N·m to meet the load requirements. For maximum speed calculation:

[0119]

[0120]

[0121] Therefore, the motor output shaft should provide a speed of about 20 rpm after deceleration, while the MG531 speed is 293 rpm. According to calculations, the MG531 geared DC motor (12V 30:1) can meet the usage requirements.

[0122] Strength check of gantry-type two-axis crossbeam

[0123] The gantry structure of the spraying device uses a transverse load-bearing beam (i.e., the X-axis guide rail), on which the spraying slider and the Y-axis linear slide are mounted. The beam is made of aluminum alloy profile, and its mechanical properties are as follows:

[0124] elastic modulus

[0125] Yield strength

[0126] Safety factor

[0127] Effective span of beam

[0128] rectangular aluminum profile

[0129] Calculation of maximum bending moment and deflection of the beam

[0130] The total mass of the spraying module, guide rail slider, spray head, etc. is approximately All loads are concentrated at the midpoint of the beam, and it is considered as a simply supported beam with a concentrated load at the midpoint. Gravity load:

[0131]

[0132] The maximum bending moment occurs at the midpoint of the beam:

[0133]

[0134] Moment of inertia of rectangular section:

[0135]

[0136] Maximum deflection (concentrated force at the midpoint of a simply supported beam):

[0137]

[0138] Within the structural tolerance range of ±1mm, it meets the usage requirements.

[0139] Maximum normal stress check

[0140] The maximum normal stress in the cross section occurs at the lower edge of the middle part of the beam:

[0141]

[0142] in,

[0143]

[0144] The yield strength is 276 MPa, taking into account the safety factor. The allowable stress is:

[0145]

[0146] Obviously:

[0147]

[0148] The safety margin is extremely high, and the beam fully meets the strength requirements.

[0149] Design calculation of gantry spraying actuator

[0150] The spraying system uses a miniature water pump to draw water from a tank and atomize it through nozzles, simulating road marking operations. The required water pump flow rate is... The nozzle outlet diameter is .

[0151] The theoretical outlet velocity of the nozzle (continuous flow, Bernoulli estimate) is:

[0152]

[0153] To achieve this speed, the minimum head required, ignoring energy loss, is:

[0154]

[0155] Considering system pressure loss, the head is selected. DC water pump.

[0156] Power system and battery life analysis: Power consumption statistics are shown in the table below:

[0157] Jetson Nano 8 1 8 3 × Stepper motors 10 0.5 5 2 × Water pumps 3 0.5 1.5 STM32 + Control module 1 1 1 Driver + Light alarm 2 0.5 1 2 × Cart motors 5 1 5 Total ≈ 21.5Wh

[0158] Battery selection: Use a 12V, 4Ah lithium battery pack, with a theoretical energy of: ;

[0159] With an effective utilization rate of 80%, the available power is approximately 38Wh, sufficient for 1 to 1.5 hours of continuous operation. For extended operation, an external battery or solar charging module can be added.

[0160] Working principle and performance analysis of the present invention

[0161] This invention realizes an intelligent road marking robot based on artificial intelligence and embedded control technology, aiming to introduce a new intelligent, safe, and automated approach to traditional road marking operations. The robot adopts a modular design concept and consists of a perception system, a control system, an execution system, a human-machine interaction system, a power system, and auxiliary devices (such as a sweeping mechanism). Through collaborative work, these systems jointly complete a series of intelligent functions such as target detection, safety warning, path movement, and marking application.

[0162] After powering on, the robot enters standby mode, with the perception system being the first to activate. The system uses a Jetson Nano as its AI processing platform, equipped with an AstraS depth camera to acquire real-time images and depth information of the construction area. A YOLOv5 target detection model is deployed on the Jetson Nano to identify human targets. This model boasts high detection accuracy and processing speed; even on resource-constrained embedded platforms, after pruning and lightweight optimization, it can still run stably at 10-15 frames per second. The system continuously monitors the area in front of the camera. When it detects personnel approaching the construction area, the Jetson Nano immediately sends an alarm command to the STM32 main control board via serial port. Upon receiving the command, the STM32 controls a buzzer to sound and simultaneously triggers flashing LEDs, creating a dual visual and auditory warning mechanism to promptly alert personnel on the construction site to avoid entering the spraying path or operating area, effectively improving the safety of the construction process.

[0163] After completing the initialization of the safety detection module, the robot enters the path execution and spraying phase. The chassis is driven by two sets of DC geared motors, and the PWM wave duty cycle controlled by STM32 is used to adjust the robot's speed and steering angle. A simple sweeping device is installed at the front of the robot, with a small motor driving a rotating brush to clean up debris and dust on the ground, providing a clean working surface for subsequent spraying and preventing debris from affecting the spraying effect or clogging the nozzles.

[0164] The spraying execution section employs a gantry structure to achieve dual-axis motion control in the XY plane. Two sets of slide rails on the crossbeam are connected to X-axis and Y-axis stepper motors, respectively. The nozzles are fixed to the slides and moved along a preset trajectory within the construction area by the motors. This structure is simple, stable, and facilitates precise control of the spraying start and end points and path. During spraying, a water pump draws water from a tank mounted on the vehicle body, delivers it to the nozzles via hoses, and completes fixed-point or continuous spraying under the control of solenoid valves or switches. The nozzles can spray straight stripes or draw simple patterns, such as the "V" shape, under the control program, achieving the effect of simulating actual road marking construction.

[0165] The human-machine interface is implemented via a Bluetooth module, allowing users to control the robot's start, stop, steering, and spraying actions via Bluetooth using their smartphones. A Bluetooth communication module (such as the HC-05) connects to the STM32's serial interface. The mobile app sends control commands (such as "forward," "turn left," and "start spraying") to the robot, which then parses the commands and drives the motors and nozzles to perform the corresponding actions. The entire interaction process features low latency and simple operation, requiring no additional professional training for users to master.

[0166] In terms of performance, all parts of the system demonstrate good practicality and engineering adaptability. In the perception system, the YOLOv5 model runs stably on the Jetson Nano platform, and combined with AstraS depth information, it can accurately identify personnel targets even in complex backgrounds. The personnel detection alarm response delay is controlled within 150ms, ensuring timely warnings before personnel enter dangerous areas. The control system uses an STM32 microcontroller, whose low power consumption and high performance ensure the machine's overall real-time response capability. In the motion control section, the DC geared motor has strong load capacity and good low-speed stability, maintaining smooth operation even in complex ground environments; the stepper motor, combined with a synchronous belt structure, achieves high-precision positioning control, with the spraying path error controlled within ±1mm, meeting the requirements of general ground marking construction. The spraying system's water pump flow design is reasonable, and combined with the simple and easy-to-maintain nozzle assembly, it can stably complete various trajectory spraying tasks.

[0167] The system performance indicators that the device of the present invention needs to achieve are shown in the table below:

[0168] 1 Overall weight 20kg 2 Travel speed 0.2m / s 3 Moving speed of gantry for spraying 0.05m / s 4 Types of sprayable patterns Supports simple patterns including straight lines, oblique lines, Chinese character "Ren" and the like 5 Air pump flow rate 300 mL / min 6 Battery life ≥ 1 hour (continuous operation) 7 Overall dimensions 100cm × 72cm × 67cm

[0169] In addition, the robot's power system is powered by a 12V lithium battery pack, meeting the operational requirements of modules such as the Jetson Nano, stepper motor, water pump, and vehicle motor. Actual testing showed it can operate continuously for 1 to 1.5 hours. The system's power consumption is reasonably controlled; future development could consider adding a power management module or an external battery to extend its runtime.

[0170] This invention focuses on three core objectives: "intelligent perception, safety early warning, and precise marking." It designs and implements an intelligent road marking robot with basic automated construction capabilities. Combining multiple cutting-edge technologies, it has significant innovations in system structure and functional integration.

[0171] 1) In terms of the perception system, Jetson Nano is used in conjunction with AstraS depth cameras and YOLOv5 algorithm is deployed to realize real-time detection and early warning of personnel in the construction area. It has good construction safety assistance capabilities and introduces a "safety collaboration" mechanism to traditional unmanned construction equipment.

[0172] 2) The control system adopts the STM32 embedded platform and realizes remote control by mobile phone through Bluetooth connection. Combined with the gantry-type two-axis structure, it realizes precise movement of the nozzle and control of the spraying path. It not only supports conventional straight line drawing, but also completes basic graphics such as "human" pattern, and has good path expansion capability.

[0173] 3) A sweeping device has also been added to the front of the system to clean the construction surface before spraying, improving the clarity and adhesion of the markings, reflecting a systematic optimization of the entire operation process.

[0174] This system is suitable for scenarios such as urban road marking, campus construction, and industrial park signage, and has significant practical value, especially in small-scale, distributed, densely populated, or safety-critical construction areas. In the future, it can be expanded to include functions such as path planning, high-precision positioning, and paint control, gradually evolving into a fully autonomous operating system, demonstrating excellent application prospects and promotional potential.

[0175] References of this invention

[0176] [1] Yang Cheng, Hao Runsheng. Design of a single-chip microcomputer control system for an automatic road marking vehicle [J]. Automation Instrumentation, 2017, 38(11):43-45. DOI:10.16086

[0177] [2] Yuan Weiqi, Wang Huili. Research on detection method and detection device for road marking vehicle baseline [J]. Instrumentation Technology and Sensors, 2008, (08): 101-103.

[0178] [3] Li Hong. Research on road boundary detection and tracking of road marking machine based on single-line lidar [D]. Jiangsu University of Science and Technology, 2024. DOI:10.27171.

[0179] [4] Jiao, Shimeng. Research on Control System of Automatic Walking Road Marking Machine [D]. Taiyuan University of Technology, 2021. DOI:10.27352

[0180] [5] Wu, Jia-Chi. Research on trajectory tracking control of road marking robot based on RTK / INS integrated navigation [D]. Shanghai University, 2023. DOI:10.27300

[0181] [6] Han Yaxuan. Research on path recognition and tracking of road marking machine based on machine vision [D]. Jiangsu University of Science and Technology, 2022. DOI:10.27171.

[0182] Finally, it should be noted that the above descriptions are merely preferred embodiments of the present invention and are not intended to limit the present invention. Although the present invention has been described in detail with reference to the foregoing embodiments, those skilled in the art can still modify the technical solutions described in the foregoing embodiments or make equivalent substitutions for some of the technical features. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of the present invention should be included within the protection scope of the present invention.

Claims

1. An intelligent road marking robot integrating AI recognition and embedded control, characterized in that: include: Vehicle walking system, ground cleaning module, gantry spraying actuator, AI perception and early warning system, embedded control system, Bluetooth human-machine interaction module and power management module; The vehicle's walking system adopts a four-wheel differential drive structure, including a frame, drive wheels and follower wheels mounted on the frame, and a DC geared motor that drives the drive wheels; The ground cleaning module is located at the front of the vehicle frame and includes a cleaning frame, a detachable roller brush mounted on the cleaning frame, and multiple DC cleaning motors that drive the detachable roller brush. The multiple DC cleaning motors drive the detachable roller brush to clean dust and debris from the road surface, and the cleaning action is linked to the vehicle's walking system for starting and stopping. The gantry-type spraying actuator is installed in the middle of the vehicle frame and includes a gantry frame, an atomizing nozzle mounted on the gantry frame, a micro water pump, a water tank, and a leak-proof device. The gantry frame includes an X-axis slide rail, a Y-axis slide rail, an X-axis stepper motor, a Y-axis stepper motor, and a ball screw. The X-axis slide rail is arranged laterally, and the Y-axis slide rail is slidably connected to the X-axis slide rail. The atomizing nozzle is slidably connected to the Y-axis slide rail. The X-axis stepper motor and the Y-axis stepper motor drive the atomizing nozzle to move in a two-dimensional plane through the ball screw. The micro water pump draws pigment from the water tank and sprays it through the atomizing nozzle. The leak-proof device enables immediate shutdown. The AI ​​perception and early warning system is installed on the upper part of the vehicle frame and includes a JetsonNano edge computing platform, an AstraS depth camera, and an audible and visual alarm unit. The AstraS depth camera collects color images and depth information of the construction area in real time. The JetsonNano deploys a lightweight YOLOv5 target detection model to identify personnel targets in real time. When personnel enter the preset safe area, the audible and visual alarm unit is triggered, and the early warning response delay is less than or equal to 500ms. The embedded control system uses an STM32 microcontroller as its underlying core, forming a dual-master architecture with a Jetson Nano. The STM32 is responsible for motion control and logic scheduling for walking, sweeping, and spraying; the Jetson Nano is responsible for image recognition and early warning decisions. The two communicate via serial port and use CRC check to ensure data reliability.

2. The intelligent road marking robot integrating AI recognition and embedded control according to claim 1, characterized in that: The vehicle frame is made of aluminum alloy profile, and the detachable roller brush is made of flexible nylon. The DC geared motor is controlled by an embedded control system to realize forward, reverse, steering and constant speed cruise, with a travel speed of 0.2m / s.

3. The intelligent road marking robot integrating AI recognition and embedded control according to claim 1, characterized in that: The atomizing nozzle can spray markings with a width of 5-8cm, with a spraying movement speed of 0.05m / s. It supports spraying straight lines, diagonal lines, and "human" shaped patterns, and the spraying trajectory error is less than or equal to ±1mm.

4. The intelligent road marking robot integrating AI recognition and embedded control according to claim 1, characterized in that: The AstraS depth camera has an effective recognition distance of 0.5 to 3 meters and a personnel recognition accuracy of no less than 95%.

5. The intelligent road marking robot integrating AI recognition and embedded control according to claim 1, characterized in that: The robot measures 100cm × 72cm × 67cm and weighs less than or equal to 20kg.

6. The intelligent road marking robot integrating AI recognition and embedded control according to claim 1, characterized in that: The Bluetooth human-machine interaction module uses an HC-05 Bluetooth module, which connects to the STM32 to receive commands from a mobile APP, enabling remote start / stop, trajectory setting, spraying control, and status feedback.

7. The intelligent road marking robot integrating AI recognition and embedded control according to claim 1, characterized in that: The power management module uses dual independent power supplies: a 5V lithium battery pack powers the Jetson Nano, and a 12V lithium battery pack powers the STM32, motor, water pump, and alarm unit, providing continuous operation for no less than 1 hour.

8. The intelligent road marking robot integrating AI recognition and embedded control according to claim 1, characterized in that: The machine uses a differential-driven MG513 motor as the main walking mechanism. The total mass of the machine is M=20kg, the rolling friction coefficient is μ=0.02, the wheel radius is r=0.09m, and the expected maximum speed is v=0.2m / s. According to the dynamic formula: Because the driving force needs to be greater than the rolling friction, to ensure a starting margin, the required driving force is calculated as 1.5 times: The required motor output torque is: The MG531 motor output shaft reduction ratio is i=30:1 for torque increase. Therefore, the rated torque of the DC geared motor should be between 0.3 and 0.5 N·m to meet the load requirements. For maximum speed calculation: The motor output shaft should provide a speed of about 20 rpm after deceleration, while the MG531 speed is 293 rpm. According to calculations, the MG531 geared DC motor (12V 30:1) can meet the usage requirements.

9. The intelligent road marking robot integrating AI recognition and embedded control according to claim 1, characterized in that, The gantry-type spraying actuator uses a transverse load-bearing beam, namely the X-axis guide rail, on which the spraying slider and the Y-axis linear slide are mounted. The beam is made of aluminum alloy profile, and its mechanical performance parameters are as follows: elastic modulus Yield strength Safety factor Effective span of beam rectangular aluminum profile The total mass of the spraying module, guide rail slider, spray head, etc. is approximately All loads are concentrated at the midpoint of the beam, and it is considered as a simply supported beam with a concentrated load at the midpoint. Gravity load: The maximum bending moment occurs at the midpoint of the beam: Moment of inertia of rectangular section: Maximum deflection (concentrated force at the midpoint of a simply supported beam): Within the structural tolerance range of ±1mm, it meets the usage requirements; The maximum normal stress in the cross section occurs at the lower edge of the middle part of the beam: in, The yield strength is 276 MPa, taking into account the safety factor. The allowable stress is: Obviously: The safety margin is extremely high, and the beam fully meets the strength requirements.

10. The intelligent road marking robot integrating AI recognition and embedded control according to claim 1, characterized in that, The gantry-type spraying actuator uses a miniature water pump to draw water from a tank and atomize it through a nozzle, simulating road marking operations. The required water pump flow rate is... The nozzle outlet diameter is ; The theoretical outlet velocity of the nozzle (continuous flow, Bernoulli estimate) is: To achieve this speed, the minimum head required, ignoring energy loss, is: Considering system pressure loss, select the appropriate head. DC water pump.

Images