Intelligent cleaning device and method for comb plate of escalator

Abstract

The application discloses an escalator comb plate intelligent cleaning device and method, and belongs to the field of intelligent cleaning robots.The device comprises an omnidirectional mobile chassis, a PLA+3D printer frame, a liftable roller brush cleaning device, a MaixCAM machine vision, a plurality of sensors, a multi-path layered power supply and an STM32 main control module; the ABBA layout of the Mecanum wheel realizes omnidirectional movement in a narrow space, the bus servo-gear pair drives the roller brush to extend into the comb gap for cleaning, and a detachable garbage recycling bin is arranged; the YOLOv8 model is carried to identify people flow, foreign matters and track the boundary of the steps, a nine-axis gyroscope is combined with a PID algorithm to correct the track, and a laser radar is used for obstacle avoidance; the multi-path independent power supply is used for isolating power and weak current interference.The application autonomously determines the cleaning time according to visual perception, replaces high-risk manual operation, and solves the problems of safety hazards, low efficiency, difficult gap cleaning, walking deviation and the like in manual cleaning; the garbage recognition rate of the prototype in an environment with sufficient light is 95.55%, the success rate of cleaning the paper towel in a lump is 96.6%, the modular structure is convenient for assembly, and the device is suitable for normal cleaning of the comb plate of escalators in places such as shopping malls and subways.

Description

Technical Field

[0001] This invention belongs to the field of intelligent cleaning robot technology for public facilities, specifically relating to an omnidirectional mobile cleaning robot suitable for autonomously cleaning foreign objects from the gaps in the comb plates at the entrances and exits of escalators. Background Technology

[0002] The cleanliness of the comb plate is a core control indicator for the safe operation and maintenance of escalators. According to GB 16899-2011 "Safety Specifications for the Manufacturing and Installation of Escalators and Moving Walks", the meshing gap between the comb teeth and the steps must not exceed 4mm, and no foreign objects should accumulate in the gap. When cigarette butts, tissues, hair, and other debris accumulate in the gaps of the comb plate, it will cause meshing failure and trigger the safety protection switch to stop the machine. From a safety perspective, the probability of a step jamming accident caused by foreign objects increases by 2.8 times. From an operation and maintenance cost perspective, comb plate wear, door lock failures, and other related maintenance failures account for 23% of the total annual failures of escalators, with an average annual additional maintenance cost of about 8,000 yuan per escalator. At the same time, manual cleaning requires cleaning staff to work in the area of ​​the running steps, and their limbs and clothing are easily caught in the transmission mechanism, posing a significant personal safety risk. Moreover, manual cleaning can only clean surface debris, and it is difficult to clean deep gaps. A complete cleaning of the escalator comb teeth takes more than 30 minutes, and there are obvious shortcomings in terms of intelligence and operation efficiency.

[0003] Application publication number CN106963295B, "An Automated Cleaning Robot for Step-Type Escalator Grooves," represents a typical solution for existing mobile escalator cleaning. This device relies on ordinary directional rollers for movement and lacks omnidirectional mobility. In the narrow spaces at escalator entrances and exits, it cannot move laterally or turn in place, making it difficult to precisely fit into the narrow gaps between the escalator steps. It is equipped with only simple infrared sensors and lacks an AI visual recognition system, so it cannot automatically identify garbage or count pedestrian flow, relying entirely on manual remote control for start and stop. Without a gyroscope attitude closed-loop correction module, roller slippage and brush reaction force during cleaning can easily cause the entire machine to deviate from the cleaning area. Furthermore, the cleaning mechanism lacks a lifting swing arm, and the brush continuously scrapes the surface of the steps during movement, easily causing damage to the paint on the steps.

[0004] Application publication number CN113003369A, "An Escalator Stairway Cleaning Device," is a fixed cleaning device for escalators, which needs to be pre-installed on the escalator body and cannot be used across multiple escalators. It adopts a wet spray cleaning scheme, which makes it easy for sewage to seep into the metal components of the escalator and cause corrosion. It does not have a dry roller brush scraping and recycling structure adapted to the narrow gaps of the comb teeth. The power supply and sensing circuit of the whole machine share a power supply circuit. The current fluctuations generated by the motor starting and stopping interfere with the recognition accuracy of the sensor. It does not have intelligent logic to determine the cleaning time autonomously and requires manual operation throughout the process.

[0005] Therefore, how to design a dedicated comb-plate automatic cleaning device that can move flexibly in all directions, is equipped with AI vision to autonomously identify debris and people, has a lifting and cleaning mechanism, real-time posture correction, and multi-circuit isolated power supply, while avoiding the accuracy attenuation caused by vibration and current interference, and achieving unmanned autonomous routine cleaning, is a technical problem that urgently needs to be solved in the current escalator operation and maintenance field. Summary of the Invention

[0006] An automatic cleaning machine for comb plates of escalators includes an omnidirectional mobile chassis module, a main support frame module, a liftable roller brush cleaning module, a machine vision perception module, a multi-sensor detection module, a layered independent power supply module, and an STM32 main control module.

[0007] The omnidirectional mobile chassis module includes a 7mm thick acrylic chassis, four sets of 42 closed-loop stepper motors, and four sets of Mecanum wheels. The Mecanum wheels adopt an ABBA square layout, with A-type wheels on the front left and rear right, and B-type wheels on the front right and rear left. The stepper motors have built-in 16384-line magnetic encoders and FOC vector drive boards, which are directly connected to the main control board via ribbon cables. The chassis has reserved motor mounting holes and frame assembly holes. The four wheels realize omnidirectional motion, including longitudinal straight movement, lateral translation, stationary rotation, and diagonal movement, based on the forward and inverse kinematics model of Mecanum wheels.

[0008] The main support frame is an FDM 3D printed PLA+ integrated structure with a 50% infill rate and a layer thickness of 0.2mm. The frame has a main control circuit board mounting slot in the middle and is threaded to the acrylic chassis at the bottom. Mirror vision platform assembly holes are symmetrically opened on both sides of the frame, and a coaxial mounting position for the cleaning mechanism is reserved at the front. Wiring harness channels are reserved inside the frame to separate the power area from the low-voltage sensing area and avoid electromagnetic interference. All assembly surfaces of the frame are vertically printed with layer patterns, and a 0.2~0.25mm assembly gap is reserved at tight fit positions. A 0.25~0.4mm gap is reserved on one side for the movable hole shaft fit.

[0009] The liftable roller brush cleaning module includes a bus servo motor, a servo disc, a square column connecting shaft, a roller brush connecting arm, a driven gear, a cleaning roller brush, and a waste collection bin. The bus servo motor is an STS5420M multi-turn bus servo motor, connected in series to the main controller via a TTL bus. The hexagonal interface of the servo disc is circumferentially limited to the square column connecting shaft, and the square column connecting shaft is anti-rotationally assembled with the square groove of the roller brush connecting arm. The other end of the roller brush connecting arm is equipped with the roller brush assembly via a bearing and a threaded sleeve. The roller brush connecting shaft is a high-load-bearing printed part with a 90% fill rate, and one end is embedded in the cleaning roller brush. The inner hole of the roller brush has a driven gear fixed at the other end; the output end of the servo motor is equipped with a driving gear, which meshes with the driven gear to drive the cleaning roller brush to rotate; the cleaning roller brush is 100mm long, 90mm in outer diameter, and has 30mm long bristles; the inner side of the roller brush shell is provided with a debris-removing protrusion, which removes foreign objects from the gaps to the detachable debris collection bin below when the cleaning rotates; the bus servo motor drives the roller brush connecting arm to complete the lifting and lowering actions: when the robot moves with the steps, the roller brush is raised to avoid scratching the steps, and after reaching the comb plate area, the roller brush is lowered to clean the gaps between the comb teeth.

[0010] The machine vision perception module uses the MaixCAM vision unit, which is symmetrically installed on both sides of the rack via mirror vision connectors. The MaixCAM is equipped with an SG2002 heterogeneous multi-core AI chip, a 1TOPS computing power NPU, and a built-in YOLOv8 recognition model trained and deployed by MaixHub.

[0011] The multi-sensor detection module includes a JY901S nine-axis attitude sensor and a TF03 lidar. The JY901S nine-axis attitude sensor is connected to the main controller via a serial port and outputs the fuselage heading angle and pitch angle data in real time for closed-loop correction of motion deviation. The TF03 lidar is powered independently with 5V, has a range of 0.1~100m, and detects the distance between the fuselage and the comb plate and steps to achieve boundary obstacle avoidance.

[0012] The power supply module inputs a 12V DC power supply and integrates four LM2596s DC-DC step-down units. The four independent outputs do not interfere with each other: 6.5V output to the bus servo motor, 6V backup servo motor power supply, 5V to the MaixCAM vision module (with an independent on / off switch), and 5V to the STM32 main controller and sensors.

[0013] The main control board uses two TPRT9013-33GB LDO linear regulators to convert 5V to 3.3V, providing independent power supplies for the main control chip and attitude sensor, and isolating power load noise. The main control circuit is equipped with an 8MHz active crystal oscillator, RC reset circuit, and SWD four-wire debugging interface. The PCB adopts a four-layer board design, with complete ground plane and power plane to suppress pulse interference.

[0014] The main control chip is an STM32F407VET6, which has two built-in core control algorithms:

[0015] (1) Stepper motor trapezoidal acceleration and deceleration algorithm: Divide the motor motion into three stages: uniform acceleration, uniform speed and uniform deceleration, smoothly adjust the pulse frequency, and eliminate the problems of frequent start-stop step loss and machine body vibration.

[0016] (2) Gyroscope PID motion offset correction algorithm: The main controller reads the heading angle of JY901S, calculates the error between the actual heading and the target heading, outputs the compensation speed through PID calculation, corrects the speed of the four Mecanum wheels in real time, and offsets the trajectory offset caused by the reaction force of the roller brush and the ground slippage.

[0017] The main controller receives garbage coordinates, tier offset data, and lidar ranging data from the vision module, autonomously plans the movement path, controls the lowering height of the roller brush, and adjusts the roller brush speed to complete fixed-point cleaning.

[0018] Further optimization of the scheme: the 3D printed parts uniformly adopt FDM process with a layer thickness of 0.2mm; the infill rate of structural load-bearing parts is ≥50%, the infill rate of vision connector and driven gear is ≥40%, and the infill rate of roller brush connecting shaft is ≥90%; after printing, the supports are removed and burrs are sanded off, and the parts are cleaned and dried before assembly; the tolerance of unspecified dimensions is uniformly ±0.2mm; the parallelism error of the two drive wheel axes is ≤0.5°, and the single-sided clearance between the wheel axle and the motor output shaft is 0.1~0.2mm; after assembly, all moving parts are manually rotated without jamming, and there is no abnormal noise or deviation after 3~5 minutes of no-load power-on.

[0019] The waste recycling bin adopts a pull-out, detachable structure with limit buckles on the side of the bin. After cleaning, it can be pulled out directly to empty foreign objects without disassembling the entire machine.

[0020] The MaixCAM vision module integrates a 2.3-inch touchscreen, which displays the recognition screen, people count, and cleaning status in real time during the debugging phase, and supports WiFi wireless remote screen transmission.

[0021] The main control board has 6 reserved serial ports, 2 IIC ports, 6 sets of step pulse interfaces, and 3 PWM output interfaces, as well as 4 reserved physical buttons for easy function expansion and on-site debugging.

[0022] The present invention provides a method for cleaning elevator comb plates, comprising the following steps:

[0023] S1. The device is powered on, the power supply module outputs four-channel regulated voltage, the main control is initialized, and the JY901S magnetic field calibration, stepper motor parameter configuration, and MaixCAM model loading are completed.

[0024] S2 and MaixCAM vision modules capture escalator images in real time and identify pedestrian flow: when pedestrian flow exceeds the threshold, the entire machine stands still while the roller brush remains raised; when pedestrian flow is below the threshold, the cleaning process begins.

[0025] S3, the vision module tracks the yellow boundary of the steps through line and color recognition, and the main control unit combines the lidar ranging data to control the omnidirectional chassis to move to the comb plate working position;

[0026] S4. The bus servo motor drives the roller brush connecting arm to lower the cleaning roller brush. The gear transmission mechanism drives the roller brush to rotate. The brush bristles extend into the gaps between the comb teeth to clean foreign objects. The foreign objects are peeled off by the protrusions on the roller brush shell and fall into the waste recycling bin.

[0027] S5. During the cleaning process, the JY901S collects the body posture in real time, and the PID offset correction algorithm dynamically adjusts the speed of the four wheels to ensure that the body is parallel and in contact with the comb plate; the vision module continuously identifies the debris, and if no foreign object is detected, it moves to the next cleaning point;

[0028] S6. After cleaning the entire comb plate, the bus servo motor lifts the roller brush, and the omnidirectional chassis moves away from the comb area, waiting for the next cleaning instruction; when the waste recycling bin is full, the equipment issues a prompt signal, and the bin is manually removed for cleaning.

[0029] Compared with existing technologies, the beneficial effects of this invention are as follows: This equipment completely avoids the safety risks of entanglement and falls during manual escalator cleaning by using fully automatic unmanned operation and real-time obstacle avoidance with lidar; it achieves excellent gap cleaning effect by relying on a liftable reverse roller brush structure and an independent waste recycling bin; it adopts a four-wheel Mecanum wheel layout and trapezoidal acceleration and deceleration algorithm to adapt to operation in narrow spaces; it integrates multi-source sensing technology to achieve fully autonomous intelligent cleaning; and it combines layered independent power supply and modular assembly design to ensure circuit anti-interference and maintenance convenience. Standardized 3D printing process and stable control system ensure structural accuracy and operational reliability. Overall, it significantly improves the efficiency and safety of escalator cleaning operations, reduces labor costs and operational hazards, and has good application benefits. Attached Figure Description

[0030] The accompanying drawings are provided to further understand the technical solutions of the present invention and constitute a part of the specification. They are used together with the embodiments of the present invention to explain the technical solutions of the present invention, and do not constitute a limitation on the technical solutions of the present invention.

[0031] Figure 1 This is a schematic diagram of the overall structure of the present invention;

[0032] Figure 2 These are the front sectional view and side sectional view of the present invention;

[0033] Figure 3 Schematic diagram of the installation model of the liftable roller brush;

[0034] Figure 4 A schematic diagram of the model structure for the vision module;

[0035] Figure 5 This is the control logic flowchart of the main control system. Detailed Implementation

[0036] 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.

[0037] To make the objectives, technical solutions, and advantages of this invention clearer, the invention will be further described in detail below with reference to the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are merely illustrative and not intended to limit the invention.

[0038] Example 1: Assembly of the overall mechanical structure

[0039] 1. Fabrication of the main load-bearing module

[0040] Non-standard parts such as the fuselage frame, vision platform connectors, roller brush connecting arms, and gear seats are printed using PLA+ material via FDM 3D printing with a layer thickness of 0.2mm. The load-bearing structure has a 65% infill rate, while the outer shell decorative parts have a 30% infill rate. After printing, the supports are removed, the assembly surfaces are sanded, and a 0.22mm gap is reserved for the tight-fitting positioning holes, and a 0.3mm gap is reserved on each side for the movable shaft holes. The base plate is made of 7mm acrylic laser-cut, with pre-drilled mounting holes for the motor, circuit board, and vision platform. The main control four-layer PCB board is fixed in the middle of the fuselage frame, mirror vision connectors are symmetrically assembled on both sides, and a coaxial mounting base for the cleaning servo is reserved at the front.

[0041] 2. Omnidirectional moving chassis assembly

[0042] Four 42 closed-loop stepper motors are fixed to the four corners of an acrylic base plate by brackets, with the motor output shafts connected to Mecanum wheels. The wheel system adopts an ABBA square layout, with A-type Mecanum wheels on the front left and rear right, and B-type wheels on the front right and rear left. After assembly, the four wheels can be manually rotated without jamming, and the parallelism error of the four wheel axles is controlled to ≤0.5°. The clearance between the motor shaft and the wheel hub on one side is 0.15mm. After assembly, the system is powered on and run under no-load for 4 minutes without deviation or abnormal noise. The main controller drives the stepper motors by outputting pulses through a timer, and has a built-in trapezoidal acceleration and deceleration algorithm, resulting in smooth start and stop without jitter.

[0043] 3. Assembly of the adaptive roller brush cleaning mechanism

[0044] The STS5420M bus servo is fixed to the front base of the fuselage. The servo disk is interference-fitted with the hexagonal connecting shaft and reinforced with screws. The connecting shaft is inserted into the square groove of the roller brush connecting arm to achieve circumferential positioning. The roller brush is embedded in the connecting shaft at both ends and equipped with miniature bearings to reduce friction. The driven gear is installed at the active end of the roller brush and meshes with the active gear of the servo output in a 1:1 ratio. The roller brush is encased in a shell, with an integrally formed baffle protrusion on the inner side of the shell. The bottom of the shell has a snap-on removable waste recycling bin.

[0045] Working mode: When the robot moves on the steps, the servo motor drives the connecting arm to lift the roller brush off the ground to avoid scratching the steps; after reaching the comb plate working position, the servo motor lowers the roller brush to fit into the gap of the comb teeth, and the servo motor drives the roller brush to rotate clockwise in multi-turn mode. The brush bristles scrape out cigarette butts and tissues in the gap in the opposite direction, and they are peeled off by the material blocking protrusion and fall into the trash can.

[0046] Example 2: Control System Hardware Implementation

[0047] 1. Main control circuit PCB design

[0048] The main controller uses an STM32F407VET6, a four-layer PCB with a top signal layer, inner power layer, inner ground layer, and bottom signal layer. An 8MHz active crystal oscillator is placed adjacent to the MCU's OSC_IN pin, minimizing trace length. The RC reset circuit uses a 100kΩ resistor and a 100nF capacitor, providing a 10ms power-on reset delay. Two TPRT9013-33GB LDOs independently output 3.3V, one supplying the main controller and the other supplying the JY901S attitude sensor; the two power supplies are isolated to reduce analog sensor noise. A 4-pin SWD download interface is onboard, with the BOOT0 pin pulled down to ground by 10kΩ, supporting online program debugging and burning.

[0049] 2. Multiple independent power supply systems

[0050] The entire unit is powered by a 12V lithium battery DC input, with four independent LM2596s DC-DC step-down modules: ① 6.5V output to the bus servo motor; ② 6V output to general peripherals; ③ 5V output to the MaixCAM vision module, with a separate boat-shaped switch on each branch; ④ 5V output to the STM32 main control board. Each DC-DC input has a 10μF + 1μF filter capacitor, and the output has a 1μF + 100nF ceramic capacitor. Wide copper traces and heat dissipation vias ensure low temperature rise under high current conditions.

[0051] 3. Peripheral hardware wiring

[0052] The JY901S nine-axis attitude sensor is connected to the main controller USART1 via serial port, outputting attitude data at a high frequency of 200Hz; the MaixCAM vision module is connected to the main controller USART2 via serial port, transmitting and identifying coordinates, categories, and offset angles; the four-channel stepper motor driver board is connected to the main controller's timer pulse IO; multiple STS5420M bus servo motors are connected in series via TTL serial ports to the main controller USART3, enabling multi-servo motor addressing control via a single serial port; the TF03 laser radar is connected to the main controller USART4 via serial port, used for escalator edge ranging and obstacle avoidance.

[0053] Example 3: Implementation of Software Algorithm Control Flow

[0054] 1. Stepper Motor Trapezoidal Acceleration and Deceleration Algorithm

[0055] The main controller internally stores a trapezoidal acceleration and deceleration sequence, which is divided into three segments: uniform acceleration, uniform speed, and uniform deceleration. At startup, a low-frequency pulse is used to start the machine, and the pulse frequency is linearly increased to the target speed. After approaching the target position, the pulse frequency is linearly reduced to stop the machine, eliminating step loss and jitter of the stepper motor, and adapting to the frequent start-stop cleaning conditions of the robot.

[0056] 2. PID Closed-Loop Correction Algorithm for Heading Misalignment

[0057] The main controller reads the heading angle after fusion by JY901S in real time and calculates the error e(t) between the actual heading and the target heading. It uses proportional-integral-derivative PID to output the compensation control quantity and substitutes it into the Mecanum wheel inverse kinematics formula to adjust the speed of the four wheels in real time. If the heading change exceeds the threshold, it starts feedforward compensation to quickly correct the deviation, offset the trajectory deviation caused by the rotation reaction force of the roller brush and the ground slippage, and ensure that the fuselage is always parallel to the comb plate.

[0058] 3. Machine vision-linked cleaning logic

[0059] 4. MaixCAM runs three algorithms simultaneously:

[0060] ① YOLOv8-Pose human key point detection, ByteTracker trajectory tracking and statistics of escalator entrance and exit traffic. If the traffic exceeds the threshold, cleaning will be paused and cleaning will be started when the traffic is sparse.

[0061] ② Based on the cigarette butt and tissue target detection model trained by MaixHub, after identifying foreign objects, the robot sends a cleaning command to the main controller, and the robot cleans repeatedly in fixed points until there are no foreign objects identified.

[0062] ③ Color threshold segmentation extracts the yellow warning zone of the steps, Hough line detection obtains the step boundary angle, calculates the lateral offset and transmits it to the chassis motion control module, and automatically adjusts the body position to align with the comb plate.

Claims

1. An automatic cleaning machine for escalator comb plates, characterized in that, The system includes a frame module, an omnidirectional walking chassis module, a liftable roller brush cleaning module, a vision perception module, a layered power supply module, and an STM32 main control module. The frame module consists of the base (3) in the attached diagram, various printing connectors, and bearing components. The omnidirectional walking chassis module includes the robot chassis (1), stepper fixing component (2), and stepper motor connecting sleeve (29). The liftable roller brush cleaning module includes the servo drive components (4), (11), and (26) in the attached diagram. 6) Roller brush connecting arm, (7)(10) gear pair, (8) roller brush, (13)(24) bearing, (14)(20) roller brush shaft; The visual perception module includes the mirror connection bracket in Figure (5)(25), (12) vision platform, (16) MaixCAM, (17)(18) angle adjustment component; The layered power supply module is the copper column in Figure (21), (23) power board; The whole machine relies on the built-in algorithm of the main control to realize omnidirectional movement, visual recognition, trajectory correction and automatic cleaning.

2. The automatic cleaning machine for escalator comb plates according to claim 1, characterized in that, The base (3), vision connector, and roller brush connecting arm are all PLA FDM printed parts with a layer thickness of 0.2mm; the load-bearing structure has a fill rate of ≥50%, and the outer shell has a fill rate of 20%~30%; the tight fit has a reserved gap of 0.2~0.25mm, the single-sided gap of the movable shaft is 0.25~0.4mm; the single-sided gap between the wheel axle and the sleeve (29) is 0.1~0.2mm, the parallelism of the drive wheel axis is ≤0.5°, and the unspecified dimensional tolerance is ±0.2mm.

3. The automatic cleaning machine for escalator comb plates according to claim 1, characterized in that, The roller brush (8) is 100mm long, 90mm in outer diameter, and has 30mm bristle length; the roller brush connecting shaft (14) is provided with anti-rotation protrusions, and the roller brush connecting arm (6) has built-in garbage stripping protrusions, which work in conjunction with the gear pair (7) (10) to complete the cleaning of gaps.

4. The automatic cleaning machine for escalator comb plates according to claim 1, characterized in that, The MaixCAM (16) is equipped with a YOLOv8 recognition model. It is installed and adjusted by relying on the vision platform (12), mirror bracket (5)(25), small servo motor (17), and connecting shaft (18). It can identify people flow, cigarette butts, and tissues and track the ladder boundary. It can also exchange data with the main controller via serial port.

5. The automatic cleaning machine for escalator comb plates according to claim 1, characterized in that, The bus servo (11) is equipped with an aluminum alloy servo disk (4) and a hexagonal connector (26) to transmit torque; in single-turn mode, it drives the roller brush connecting arm (6) to lift and lower, and in multi-turn mode, it drives the gear pair to rotate the roller brush, and has real-time feedback protection for load and temperature.

6. The automatic cleaning machine for escalator comb plates according to claim 1, characterized in that, The large bearing (13) is assembled at both ends of the roller brush, and the small bearing (24) is assembled on the vision shaft to reduce rotational friction. After all moving parts are assembled, they can be manually rotated without jamming.

7. The automatic cleaning machine for escalator comb plates according to claim 1, characterized in that, The power board (23) is suspended above the base (3) of the whole machine by copper pillars (21), and integrates four independent LM2596 step-down units. The visual power supply branch is equipped with an independent switch, and the main control is equipped with dual 3.3V LDOs and multi-stage filter capacitors.

8. The automatic cleaning machine for escalator comb plates according to claim 1, characterized in that, The robot chassis (1) is made of acrylic sheet, and the stepper fixing part (2) locks the 42 closed-loop stepper motor. The stepper motor connecting sleeve (29) connects the motor output shaft and the Mecanum wheel. The chassis adopts an ABBA four-wheel layout, and the main control has a built-in trapezoidal acceleration and deceleration algorithm to realize three degrees of freedom omnidirectional walking.

9. The automatic cleaning machine for escalator comb plates according to claim 1, characterized in that, The machine is equipped with a JY901 gyroscope and the main controller has a built-in PID offset correction algorithm to correct the speed of the four wheels in real time and counteract the trajectory deviation caused by the sweeping reaction force and ground slippage.

10. An automatic cleaning machine for escalator comb plates according to claim 1, characterized in that, The automatic cleaning control steps for the entire machine are as follows: S1. When the device is powered on, the power board (23) outputs multiple regulated voltages, and the main controller, maixcam (16), and gyroscope are initialized and calibrated. S2, maixcam (16) identifies the flow of people. When the flow of people exceeds the set threshold, the servo motor (11) drives the roller brush connecting arm (6) to lift the roller brush (8), and the whole machine goes into standby mode. S3. After the flow of people reaches the target, maixcam (16) identifies the yellow boundary of the ladder and outputs the offset. The main controller controls the Mecanum wheel mounted on the robot chassis (1) to move omnidirectionally to the comb plate working position. S4, the servo motor (11) drives the roller brush connecting arm (6) to lower the roller brush (8), and the cylindrical gear (10) meshes with the driven gear (7) to drive the roller brush (8) to rotate and clean the foreign objects in the gaps; The S5 and JY901 gyroscopes provide real-time feedback on the fuselage heading, and the main control PID algorithm dynamically adjusts the speed of the four wheels to counteract the trajectory deviation caused by the rotational reaction force of the roller brush. After S6 and maixcam (16) receive no garbage identification signal, the servo motor raises the roller brush, the chassis moves away from the comb area, and waits for the next cleaning instruction.

Images