A wireless charging control and guidance system and method for automated guided vehicles.

Abstract

This invention discloses a wireless charging control and guidance system for automated guided vehicles (AGVs), belonging to the field of wireless charging technology for AGVs. The system includes: a charging base station with a travel mechanism movable along the X-Y-Z directions and a wireless transmitting module at its end; a binocular vision perception module fixedly installed above the base station for acquiring stereo images of the receiving module at the bottom of the AGV; and a pose calculation engine that calculates the six-degree-of-freedom pose deviation of the receiving module relative to the transmitting module through dense stereo matching and 3D point cloud registration. This invention replaces the traditional repeated stopping and correction mode at the vehicle end with active adaptation at the charging end, completely decoupling the centering process from the AGV scheduling system. It achieves a wireless charging efficiency of over 92% without improving AGV stopping accuracy, and does not rely on any manual ground markings. It also possesses multi-vehicle adaptive capabilities, providing a highly reliable and efficient charging solution for unmanned AGV operations.

Description

Technical Field

[0001] This invention relates to the field of wireless charging technology for Automated Guided Vehicles (AGVs), specifically to a wireless charging control and guidance system in which the charging base station actively performs spatial pose perception and adaptive centering of the transmitting module. It is particularly suitable for unmanned AGV operation scenarios where docking accuracy is limited but charging efficiency requirements are stringent. Background Technology

[0002] Automated Guided Vehicles (AGVs), as core transfer nodes in intelligent logistics and flexible manufacturing systems, rely heavily on a robust power supply system for their continuous operation capabilities. Magnetic-coupled resonant wireless charging technology, with its advantages of being contactless, wear-free, and allowing for fully enclosed installation, is gradually replacing traditional wired charging methods and becoming the ideal choice for AGV power systems.

[0003] However, wireless charging technology is extremely sensitive to the relative position between the transmitting and receiving coils. In engineering practice, AGVs are affected by multiple factors, including inherent errors in the navigation system (typically ±20mm), differences in local ground flatness, tire pressure fluctuations, and chassis deformation caused by load. Even with repeated corrections by the navigation system, there is still a significant and random spatial deviation between the final stopping position and the theoretical target position. When the center offset between the transmitting and receiving coils exceeds ±10mm, the magnetic coupling coefficient will decrease exponentially, and the charging efficiency can drop sharply from over 90% to below 60%, or even trigger system protection and prevent charging from starting.

[0004] To address this technological bottleneck, existing solutions mainly follow two paths:

[0005] Path 1: Increase coil size to improve position tolerance. While this approach can alleviate alignment pressure to some extent, it comes at a very high cost—doubling the coil area increases system volume and cost simultaneously, significantly raises the risk of electromagnetic radiation leakage, and reduces charging efficiency due to a decrease in the coil quality factor, exhibiting a clear diminishing marginal effect. This is not an economically feasible solution to the problem.

[0006] Path Two: Achieving precise positioning by visually guiding the AGV's own movement. Chinese patent CN106787266A discloses a wireless charging method for mobile robots, which sets concentric circular patterns on the surface of the transmitting coil. An onboard camera captures an image of this pattern and extracts the center position, using this image to control the robot's movement to center the coil. Chinese patent CN106712313A also employs a similar principle, printing a radial stripe pattern on the surface of the transmitting coil. An onboard camera extracts the intersection points of the center line segments of the pattern, deducing the center position of the transmitting coil and thus correcting the vehicle's posture.

[0007] While path two has achieved some success, it has the following inherent drawbacks:

[0008] First, the control mechanism is flawed. The proposed solution requires the AGV to repeatedly fine-tune its position after arriving at the charging area to "accommodate" the fixedly installed transmitting coil. This "vehicle accommodating the charging station" model deeply couples the charging alignment process with the AGV scheduling system—the vehicle cannot be released into the work queue before charging is complete, leading to a decrease in the overall logistics cycle time. Simultaneously, the fine-tuning operation consumes the AGV's own motion control resources and navigation computing power, placing an additional burden on older models or AGV controllers with limited computing power.

[0009] Second, the perception dimension is insufficient. All the above solutions use a monocular camera in conjunction with manually designed two-dimensional ground markings for positioning. Monocular vision lacks depth perception capabilities, only able to acquire the two-dimensional projection center of the transmitting coil in the image plane, and cannot perceive the three-dimensional spatial attitude of the receiving coil relative to the transmitting coil (including air gap distance deviation, pitch angle deviation, and roll angle deviation). In scenarios where the actual ground is not absolutely level or the vehicle's uneven load causes chassis tilt, the two-dimensional perception information is insufficient to fully describe the true pose relationship between the coils, resulting in an inherent information gap in centering decision-making.

[0010] Third, it relies on manual marking. This type of solution heavily depends on placing specific pattern markings on the surface of the transmitting coil, which not only increases the complexity of the manufacturing process, but also makes the markings prone to wear and contamination during long-term use, leading to visual recognition failure and raising concerns about the long-term reliability of the system.

[0011] Furthermore, Chinese patent CN109533329B discloses a wireless charging device for drones, which uses a magnetic levitation three-degree-of-freedom platform combined with binocular vision to center the drone's charging port. However, it should be noted that this solution is designed for drones hovering and landing in three-dimensional space for centering. Its motion characteristics (floating in the air), load characteristics (light load and low inertia), and environmental constraints (open outdoor space) are fundamentally different from those of AGVs docking on the ground with heavy loads. The magnetic levitation mechanism it relies on cannot support the tens to hundreds of watts of power transmission modules required for AGV charging, and it lacks the technical foundation for migration to AGV charging scenarios.

[0012] In summary, the existing technology lacks a wireless charging solution for AGVs that does not rely on repeated adjustments by the AGV itself, does not rely on manual ground markings, can obtain complete three-dimensional pose information between the transceiver coils, and achieves high-precision centering through active movement of the charging end. Summary of the Invention

[0013] To address the following core shortcomings of existing AGV wireless charging alignment technologies: first, the alignment process requires repeated movement of the AGV itself, leading to strong coupling with the scheduling system and reduced operational efficiency; second, monocular two-dimensional vision cannot perceive the three-dimensional spatial pose between the transceiver modules, resulting in incomplete alignment decision information; and third, reliance on manual ground marking leads to long-term reliability issues. The purpose of this invention is to provide a system and method where the charging base station autonomously completes spatial pose perception and active alignment of the transmitting module, achieving high-efficiency magnetic coupling docking between the transceiver coils without increasing the AGV's docking accuracy requirements.

[0014] The wireless charging control and guidance system for automated guided vehicles provided in this application adopts the following technical solution:

[0015] A wireless charging control and guidance system for an automated guided vehicle includes:

[0016] A charging base station, installed on the ground or underground in the AGV charging area, includes a fixed base, a multi-dimensional motion mechanism installed on the base, and a wireless transmission module installed at the end of the motion mechanism; the motion mechanism has translational degrees of freedom in the X and Y directions in the horizontal plane, and can selectively extend the Z-direction lifting degree of freedom, for actively adjusting the spatial position of the wireless transmission module in space.

[0017] A binocular vision perception module is fixedly mounted on a rigid bracket of the charging base station, located outside the working range of the travel motion mechanism and its end effector. It captures images of the bottom area of ​​the AGV parked in the charging area from a fixed and calibrated extrinsic posture. The binocular vision perception module includes a first industrial camera and a second industrial camera arranged at intervals along a horizontal baseline. It is used to simultaneously capture left and right eye images, including the AGV on-board receiving module, and generate a dense disparity map covering the surface area of ​​the receiving module through stereo matching.

[0018] The pose calculation engine, connected to the binocular vision perception module via a high-speed data interface signal, is configured to perform the following ordered calculation steps:

[0019] The left and right eye images are subjected to joint distortion correction, epipolar correction and image enhancement preprocessing;

[0020] Subpixel-level disparity maps are generated based on semi-global matching algorithms or neural network stereo matching methods;

[0021] Based on the principle of triangulation and the pre-calibrated intrinsic and extrinsic parameters of the binocular system, the disparity map is mapped as a three-dimensional dense point cloud on the surface area of ​​the receiving module;

[0022] The three-dimensional dense point cloud is registered in three dimensions with the pre-stored standard geometric model of the receiving module. The spatial coordinates (X_r, Y_r, Z_r) of the center point of the receiving module in the reference coordinate system of the transmitting module and the attitude angles (pitch angle α, roll angle β, yaw angle γ) of the surface normal vector of the receiving module relative to the reference direction are calculated. Thus, the six-degree-of-freedom pose deviation vector of the receiving module relative to the current reference position of the wireless transmitting module is output.

[0023] An adaptive centering controller, connected to the pose calculation engine and the drive units of each axis of the travel motion mechanism, is configured to: use the pose deviation vector output in real time by the pose calculation engine as feedback, generate motion commands for each drive axis using a visual servo control strategy, drive the travel motion mechanism to move the wireless transmission module continuously in the horizontal and / or vertical directions, and during the movement, the binocular vision perception module continuously updates the pose deviation feedback at a frequency of not less than 20Hz to form a continuous closed-loop adjustment until the comprehensive deviation between the central axis of the wireless transmission module and the central axis of the receiving module converges to within the preset centering tolerance band and remains stable for not less than a preset time.

[0024] The charging management unit is communicatively connected to the power conversion circuit of the wireless transmission module and the on-board battery management system of the AGV. It is used to automatically start the resonant charging process after the centering controller provides feedback on centering lock. During the charging process, it monitors the voltage, current, frequency, coil temperature and status parameters fed back from the battery end in real time, and dynamically adjusts the output power according to the monitoring results. When charging is completed or an abnormal state is detected, it cuts off the output and controls the stroke motion mechanism to reset to the standby initial position.

[0025] Furthermore, the travel motion mechanism is an XY two-dimensional precision slide or an XYZ three-dimensional precision slide. Each motion axis is driven by a ball screw and guided by a cross roller guide. Each axis is driven by an independent servo motor and equipped with an absolute grating encoder to form a position closed loop. The single-axis repeatability is better than ±0.05mm, the X and Y travel is not less than ±150mm, and the Z travel is not less than ±30mm, so as to cover various typical AGV docking deviation ranges.

[0026] Furthermore, both cameras of the binocular vision perception module are global shutter type industrial cameras with a baseline distance of 80mm to 200mm. The camera optical axis is installed at a fixed downward angle to the horizontal plane to ensure that the receiving module is always within the common field of view of the binocular system within the range of AGV parking height changes. Both cameras are equipped with an adjustable intensity active infrared supplementary light module to enhance the texture recognition of the receiving module surface under low illumination or bottom shadow conditions.

[0027] Furthermore, the standard geometric model of the receiving module pre-stored in the pose calculation engine is a three-dimensional template generated based on the CAD design file of the receiving module, or a reference point cloud model obtained by pre-scanning and reconstructing the receiving module under standard pose. The three-dimensional registration adopts a two-level registration strategy: first, coarse registration based on principal component analysis of the point cloud normal vector distribution, and then fine registration using the iterative nearest point algorithm, to ensure that it can still converge to the global optimal solution when the initial deviation is large.

[0028] Furthermore, the adaptive centering controller adopts a cascaded control structure consisting of an outer loop visual pose error closed loop and an inner loop motor current / speed closed loop. The outer loop controller adopts a model predictive control algorithm or a sliding mode control algorithm, and the inner loop adopts a proportional-integral-derivative control algorithm. The outer loop control cycle is synchronized with the visual sampling cycle, and the inner loop control cycle is not less than 1kHz.

[0029] Furthermore, the wireless transmitting module includes a high-frequency full-bridge inverter circuit, an LCC-type multi-stage compensation network, and a planar spiral transmitting coil connected in sequence; the receiving module includes a planar spiral receiving coil, a compensation network, a rectifier filter circuit, and a charging control circuit connected in sequence; the wireless transmitting module and the receiving module transmit energy through magnetic coupling resonance, and the system operating frequency is dynamically tuned within the range of 80kHz to 300kHz according to the real-time air gap and load status.

[0030] Furthermore, the system also includes a docking wake-up sensor group, which is at least one of an underground electromagnetic induction switch, a side-beam photoelectric sensor, or an ultrasonic ranging sensor. It is installed at the entrance and key internal locations of the charging area to continuously monitor whether the AGV has entered the effective charging range in a low-power mode. After confirming the arrival of the vehicle, it wakes up the binocular vision perception module and the subsequent system to start working, avoiding the continuous full-load operation of the system components, which would lead to a reduction in lifespan.

[0031] Furthermore, the charging management unit also integrates a foreign object and living object safety detection module, which includes an impedance spectrum analysis submodule and a visual recognition submodule. The impedance spectrum analysis submodule continuously injects weak high-frequency detection signals into the transmitting coil before charging starts and during charging, and measures its equivalent series resistance and inductance. When a parameter mutation is detected that exceeds a preset threshold, it is determined that a conductive foreign object has been introduced. The visual recognition submodule calls a pre-trained lightweight convolutional neural network to perform binary classification recognition on the image of the gap area between the transmitting and receiving coils acquired by the binocular visual perception module. The two submodules use "OR" logic output, and if either channel triggers an alarm, it will execute instantaneous interruption protection and activate an audible and visual warning.

[0032] The present invention also provides a wireless charging control guidance method based on the above system, comprising the following steps executed in sequence:

[0033] Step S1: The AGV travels to the charging area and stops according to the scheduling instructions. The docking wake-up sensor group detects that the vehicle has entered the effective range and sends a wake-up signal to the system main control unit.

[0034] Step S2: The binocular vision perception module responds to the wake-up signal and starts synchronous acquisition to obtain the left and right eye images of the area where the receiving module at the bottom of the AGV is located, and transmits them to the pose calculation engine;

[0035] Step S3: The pose calculation engine sequentially performs distortion correction, epipolar correction, stereo matching to generate a disparity map, and triangulation to reconstruct a 3D dense point cloud for the left and right images. Then, it performs two-level registration with the pre-stored standard model of the receiving module to calculate the spatial position offset (ΔX, ΔY, ΔZ) and attitude angle deviation (Δα, Δβ, Δγ) of the receiving module relative to the reference coordinate system of the transmitting module, forming a six-dimensional deviation vector.

[0036] Step S4: The adaptive centering controller receives the six-dimensional deviation vector, inputs it into the outer loop visual servo controller, calculates the speed or position target command of each axis of the stroke motion mechanism, and drives the servo motors of each axis to move the wireless transmission module toward the target centering position in the horizontal and / or vertical planes; throughout the entire movement, steps S2 to S4 are continuously executed at a cycle rate of not less than 20Hz to form a continuous visual servo closed-loop correction.

[0037] Step S5: When the absolute values ​​of ΔX, ΔY, and ΔZ in the six-dimensional deviation vector are all less than the first preset threshold (preferably ΔX≤±1.5mm, ΔY≤±1.5mm, ΔZ≤±2mm) and the absolute values ​​of Δα, Δβ, and Δγ are all less than the second preset threshold (preferably each attitude angle deviation ≤±0.5°), and this state remains stable for at least 0.5 seconds, the centering controller determines that the centering is locked.

[0038] Step S6: After receiving the centering lock signal, the charging management unit initiates a communication handshake with the AGV on-board battery management system. After confirming the battery status, it gradually increases the transmission power to the set value in a soft-start mode, enters the constant current-constant voltage multi-stage charging process, and continuously monitors the transmitting end thermal parameters and battery end feedback data during the charging process, dynamically adjusting the output to adapt to the charging process.

[0039] Step S7: When the vehicle battery management system reports that the battery state of charge has reached the full charge threshold or receives a termination command, the charging management unit performs power decay until the output is zero, disconnects the power output contactor, and then sends a reset command to the travel motion mechanism to return the wireless transmission module to the initial standby position.

[0040] Step S8: The system sends a communication signal to the AGV indicating that charging is complete and the AGV is ready to leave. The AGV then autonomously drives out of the charging area, and the system switches to a low-power monitoring mode, waiting for the arrival of the next AGV to trigger the switch.

[0041] Furthermore, in step S3, if the pose calculation engine has pre-stored multiple standard models of receiving modules corresponding to multiple vehicle models, then before 3D registration, each model is tried in turn for coarse registration, and the model with the smallest registration residual is selected as the current matching model, so as to realize adaptive recognition and pose calculation in multi-vehicle mixed traffic scenarios.

[0042] Furthermore, in step S4, after the adaptive centering controller receives the six-dimensional deviation vector for the first time, if it detects that the deviation of a certain axis exceeds 80% of the travel range of that axis, it adopts a phased approach strategy: first, it quickly reduces the deviation to within 50% of the travel range in an open-loop manner, and then switches to a closed-loop visual servo for precise approach, so as to balance response speed and centering accuracy.

[0043] Compared with the prior art, the beneficial effects of the present invention are as follows:

[0044] 1. The technical approach of shifting the focus from "vehicles adapting to charging terminals" to "charging terminals adapting to vehicles" completely decouples the charging alignment process from the AGV scheduling system. AGVs do not need to sacrifice any work rhythm for charging, resulting in a qualitative improvement in the overall system efficiency.

[0045] 2. This invention fundamentally upgrades the perception dimension, moving from monocular two-dimensional marker recognition to binocular three-dimensional markerless pose perception. Existing solutions rely on monocular cameras to capture images of artificial ground markers, only acquiring the center coordinates within a two-dimensional plane, lacking depth information and pose perception capabilities. This invention employs binocular stereo vision technology, using dense stereo matching and three-dimensional point cloud reconstruction to directly acquire the complete six-degree-of-freedom pose of the receiving module, without relying on any artificial markers on the surface of the transmitting module. This technological upgrade enables the system to perceive three-dimensional spatial pose information such as vehicle chassis tilt and air gap changes, providing complete state input for high-precision alignment.

[0046] 3. Continuous visual servo closed-loop control upgrades the discrete "detection-execution-re-detection" process to continuous closed-loop adjustment. In existing technologies, deviation detection and motion correction are separate sequential steps; a detection is completed before a motion is executed, and the detection is repeated after the motion stops. This invention constructs a visual servo closed-loop that operates continuously at a frequency of no less than 20Hz, acquiring pose deviations in real time and continuously correcting the trajectory throughout the entire motion of the transmitting module. This continuous closed-loop control mode not only achieves faster centering speed (3-8 seconds in actual measurements) but also effectively suppresses motion overshoot, achieving a smooth approach in one go, which is especially important for medium to large inertial loads.

[0047] 4. The comprehensive performance leap resulting from the system's synergistic effect is achieved by setting up "binocular 3D pose perception, active travel mechanism, and continuous vision servo closed loop" as a tightly coupled technical whole, rather than simply superimposing them: the high-dimensional state information provided by binocular vision gives the active mechanism a clear target and basis for movement, while the continuous servo closed loop maximizes the value of the high-dimensional state information; the synergistic effect of the three enables the system to control the final alignment accuracy within ±1mm under the condition that the initial AGV docking deviation is as high as ±50mm, and the charging efficiency is stable at over 92%.

[0048] 5. Since the present invention does not require any visual marking patterns to be placed on the surface of the transmitting module, it completely eliminates the risk of recognition failure caused by mark wear and contamination. The system's reliability throughout its entire life cycle is significantly better than that of schemes that rely on manual marking. At the same time, the binocular sensing module is installed on a fixed bracket above the charging base station and does not move with the transmitting module. Its calibration status is stable for a long time and the maintenance requirements are extremely low.

[0049] 6. This invention achieves pose calculation by directly identifying and registering the three-dimensional geometric structure of the receiving module itself, without relying on specific manual markings. Therefore, by pre-storing standard models of receiving modules for various vehicle models, multi-vehicle adaptive design can be achieved without adding any hardware modifications. Attached Figure Description

[0050] Figure 1 This is a schematic diagram of the structure of a wireless charging control and guidance system for an automated guided vehicle according to the present invention;

[0051] Figure 2 This is a complete algorithm flowchart of the pose calculation engine in this invention, from the original image input to the six-degree-of-freedom pose deviation output;

[0052] Figure 3 This is a flowchart of a wireless charging control and guidance method for an automated guided vehicle according to the present invention;

[0053] Explanation of reference numerals in the attached figures: 100, charging base station; 110, base; 120, travel motion mechanism; 130, wireless transmission module; 200, binocular vision perception module; 300, pose calculation engine; 400, adaptive centering controller; 500, charging management unit. Detailed Implementation

[0054] 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.

[0055] Please see Figure 1-3 The present invention provides the following embodiments:

[0056] Example 1: Implementation of a Two-Dimensional Centering System

[0057] like Figure 1 As shown, this embodiment provides an AGV wireless charging control and guidance system. Its application scenario is the final assembly workshop of an automobile manufacturing plant. The AGV is a backpack-type vehicle with a rated load of 500kg. The vehicle-mounted wireless receiving module is fixedly installed at the center of the bottom of the vehicle body. The receiving coil plane is parallel to the ground and the height above the ground is 200mm±15mm (which varies with the load).

[0058] The charging base station 100 includes a cast iron base 110, which is fixed in an installation recess below the ground in the charging area, with its upper surface flush with the ground. An XY two-dimensional cross slide (stroke motion mechanism 120) is mounted on the base 110, with a lower layer of X-axis linear module (stroke ±180mm) and an upper layer of Y-axis linear module (stroke ±150mm), the two axes stacked orthogonally. A wireless transmitter module 130 is fixed to the motion slide of the Y-axis linear module. Each axis uses a C5-grade ball screw (5mm lead) paired with a THK SR-type cross roller guide, driven by an Oriental Motor AR series closed-loop stepper motor with a built-in 1000PPR incremental encoder. The driver microstepping factor is set to 16, with a theoretical resolution of 0.005mm and a repeatability of ±0.03mm. The wireless transmitting module 130 includes a planar spiral transmitting coil (200mm outer diameter, 12 turns, Litz wire wound), an E-type manganese zinc ferrite core, an LCC type compensation network, and a high-frequency full-bridge inverter board with a rated power of 1kW, all encapsulated in an engineering plastic housing. The bottom of the housing is equipped with spring-loaded floating support rollers to prevent scratching the ground during movement.

[0059] The binocular vision perception module 200 is fixed above the charging base station via an independent steel gantry bracket. Rubber shock-absorbing pads isolate the bracket from the base 110 to prevent vibrations introduced by the travel motion mechanism 120 from being transmitted to the cameras. Both the first and second cameras are Hikvision MV-CA050-10GM global shutter industrial cameras with a resolution of 2592×1944, a frame rate of 30fps, and equipped with Kowa 6mm fixed-focus industrial lenses, offering a horizontal field of view of approximately 55°. The two cameras are fixedly installed 140mm apart along the horizontal baseline, with their optical axes pointing downwards at a 15° angle to the vertical direction, both looking down at the bottom area of ​​the AGV 10. The camera module integrates a ring-shaped infrared LED fill light panel (wavelength 850nm), whose activation and brightness adjustment are automatically controlled by an ambient light sensor.

[0060] The pose calculation engine 300 and the adaptive centering controller 400 are integrated into a Beckhoff CX5140 embedded controller. This controller is equipped with an Intel Atom x7-E3950 processor (quad-core, 1.6GHz), running a dual system of TwinCAT real-time kernel and Ubuntu. The vision algorithm runs on the Ubuntu side, and the motion control runs on the TwinCAT RT side. The two exchange nanosecond-level data through shared memory.

[0061] The charging management unit 500 is an independent chassis installed in the electrical cabinet next to the charging base station. It includes a power regulation module, EMI filter circuit, communication interface board and safety protection circuit. It is connected to the inverter board of the wireless transmission module 130 through a shielded power cable and communicates with the AGV vehicle-mounted BMS through the CANopen protocol.

[0062] The system operation process in this embodiment is as follows:

[0063] AGV 10 travels to the charging area and stops according to the scheduling instructions. Its front end touches the mechanical limit block at the end of the charging area. This block integrates a microswitch, and the switch signal is sent to the embedded controller as a stop wake-up signal. The controller then triggers the binocular vision perception module 200 to acquire one image frame. The pose calculation engine 300 performs the following processing on the image (see...). Figure 2 First, distortion correction and epipolar correction are performed on the left and right images based on pre-calibrated camera intrinsic parameters (focal length, principal point, distortion coefficients) and binocular extrinsic parameters (rotation matrix, translation vector). Then, a disparity map is generated using the SGBM (semi-global block matching) algorithm implemented in OpenCV. The cost calculation uses Census transform and gradient information fusion, and disparity post-processing includes left-right consistency checks and sub-pixel interpolation. Next, a sparse 3D point cloud (approximately 50,000 points) of the receiving module region is generated based on the disparity map and calibration parameter triangulation. Finally, the point cloud is ICP-registered with the pre-stored STL model of the receiving module for this vehicle model to obtain the pose of the receiving module center relative to the camera coordinate system. This pose is then converted into a six-dimensional deviation (ΔX, ΔY, ΔZ, Δα, Δβ, Δγ) relative to the zero-point position of the transmitting module using a pre-calibrated hand-eye matrix. In this embodiment, since the ground where the AGV is parked has been hardened and the receiving module mounting reference is rigidly connected to the vehicle chassis, the ΔZ and attitude angle deviations are usually small, and the centering controller mainly focuses on ΔX and ΔY.

[0064] The adaptive alignment controller 400 takes ΔX and ΔY as inputs and uses an incremental PID algorithm (proportional coefficient Kp=0.85, integral coefficient Ki=0.12, derivative coefficient Kd=0.05, sampling period synchronized with visual frame rate 50ms) to calculate the target pulse increment of each axis stepper motor, and sends it to the driver via EtherCAT bus. After the first round of movement is completed, the system immediately acquires a new image to update the deviation and corrects the next round of movement commands. In actual testing, when the initial ΔX=+45mm and ΔY=-32mm, the deviation converged to ΔX=+0.6mm and ΔY=-0.8mm after 4 iterations (approximately 6.5 seconds in total), meeting the preset alignment threshold (±1.5mm). After alignment is locked, the charging management unit 500 starts charging, which is completed in approximately 80 minutes, with an average charging efficiency of 91.7%. After charging is completed, the slide automatically resets and the AGV departs.

[0065] Example 2: Three-dimensional adaptive centering system (including Z-axis adjustment)

[0066] This embodiment, based on Embodiment 1, expands the stroke motion mechanism 120 into an XYZ three-dimensional slide table, adding a Z-axis lifting module (stroke ±30mm, using a lead screw lifting mechanism, equipped with a brake motor to prevent falling due to power failure) to the Y-axis slide. Correspondingly, ΔZ, pitch angle α, and roll angle β are fully utilized in the six-dimensional deviation output by the pose calculation engine 300. The controller drives the Z-axis lifting based on the ΔZ value, precisely controlling the air gap distance between the transmitting coil plane and the receiving coil plane within the range of 20mm ± 2mm to obtain the optimal coupling coefficient. Simultaneously, if the calculated α or β exceeds 1°, the system will use a difference compensation strategy—superimposing a small arc trajectory in the X and Y motions—to equivalently compensate for the impact of angular deviation on the coupling area in a spatial geometric manner. This embodiment is particularly suitable for scenarios with large differences in AGV load (the chassis height difference between empty and fully loaded can reach 25mm) or where the charging area has a slope.

[0067] Example 3: Multi-vehicle adaptive charging station

[0068] This embodiment describes a scenario in a courier sorting center, where three different brands and load capacities of AGVs are mixed together, each with different receiving module dimensions, installation positions, and geometric shapes. This embodiment does not add any hardware modifications; only the 3D CAD models of the three receiving modules are pre-stored in the storage area of ​​the pose calculation engine 300. Before each registration, the engine performs rapid coarse registration (based on FPFH feature descriptors + RANSAC) sequentially with the currently reconstructed point cloud and the three templates, selecting the template with the highest in-registration point ratio (i.e., best fit) as the current matching target, and then performs ICP fine registration. The additional time for the entire template selection and fine registration is approximately 80ms, which does not affect the overall system response speed. Experimental statistics show that the system's average centering time for the three vehicle types is 5.2 seconds, with a centering success rate of 99.3% (judgment criteria: deviation converges to within the threshold and first charging start-up success rate), fully verifying the multi-vehicle adaptive capability of this invention.

[0069] Example 4: Multi-source security protection mechanism

[0070] This embodiment deeply integrates foreign object and live object detection functions into the charging management unit 500. The impedance spectrum analysis submodule generates a 50kHz to 500kHz sweep signal (amplitude 5V, injected power <0.5W, does not affect main power transmission) using a DDS frequency synthesizer. It acquires the voltage and current phasors across the transmitting coil using a high-precision ADC (24-bit) to calculate the equivalent series resistance Rs and equivalent series inductance Ls. The system records Rs0 and Ls0 as a baseline when there is no load and no foreign object during the first power-on. During pre-charging detection, if Rs changes by more than +8% relative to Rs0 or Ls changes by more than -5% relative to Ls0, it is determined that there may be a metallic foreign object. The visual recognition submodule sends the image of the coil gap area captured by the binocular system (cropped and normalized to a 96×96 pixel grayscale image) to a pre-trained MobileNetV2 classifier (the training set contains 5000 real-world images with and without various metallic foreign objects) to output the foreign object confidence score. The two alarm signals are logically ORed and then connected to the emergency stop input of the charging management unit 500. Since the system was deployed, it has successfully identified and prevented seven potential charging safety accidents caused by fallen bolts, gaskets, etc., during three months of continuous operation, verifying the effectiveness and necessity of the dual-mode detection mechanism.

[0071] Example 5: Wireless charging control guidance method based on the above system (see...) Figure 3 The method includes the following steps:

[0072] Step S1: The AGV travels to the charging area and stops according to the scheduling instructions. The docking wake-up sensor group detects that the vehicle has entered the effective range and sends a wake-up signal to the system main control unit.

[0073] Step S2: The binocular vision perception module responds to the wake-up signal and starts synchronous acquisition to obtain the left and right eye images of the area where the receiving module at the bottom of the AGV is located, and transmits them to the pose calculation engine;

[0074] Step S3: The pose calculation engine sequentially performs distortion correction, epipolar correction, stereo matching to generate a disparity map, and triangulation to reconstruct a 3D dense point cloud for the left and right images. Then, it performs two-level registration with the pre-stored standard model of the receiving module to calculate the spatial position offset (ΔX, ΔY, ΔZ) and attitude angle deviation (Δα, Δβ, Δγ) of the receiving module relative to the reference coordinate system of the transmitting module, forming a six-dimensional deviation vector.

[0075] Step S4: The adaptive centering controller receives the six-dimensional deviation vector, inputs it into the outer loop visual servo controller, calculates the speed or position target command of each axis of the stroke motion mechanism, and drives the servo motors of each axis to move the wireless transmission module toward the target centering position in the horizontal and / or vertical planes; throughout the entire movement, steps S2 to S4 are continuously executed at a cycle rate of not less than 20Hz to form a continuous visual servo closed-loop correction.

[0076] Step S5: When the absolute values ​​of ΔX, ΔY, and ΔZ in the six-dimensional deviation vector are all less than the first preset threshold (preferably ΔX≤±1.5mm, ΔY≤±1.5mm, ΔZ≤±2mm) and the absolute values ​​of Δα, Δβ, and Δγ are all less than the second preset threshold (preferably each attitude angle deviation ≤±0.5°), and this state remains stable for at least 0.5 seconds, the centering controller determines that the centering is locked.

[0077] Step S6: After receiving the centering lock signal, the charging management unit initiates a communication handshake with the AGV on-board battery management system. After confirming the battery status, it gradually increases the transmission power to the set value in a soft-start mode, enters the constant current-constant voltage multi-stage charging process, and continuously monitors the transmitting end thermal parameters and battery end feedback data during the charging process, dynamically adjusting the output to adapt to the charging process.

[0078] Step S7: When the vehicle battery management system reports that the battery state of charge has reached the full charge threshold or receives a termination command, the charging management unit performs power decay until the output is zero, disconnects the power output contactor, and then sends a reset command to the travel motion mechanism to return the wireless transmission module to the initial standby position.

[0079] Step S8: The system sends a communication signal to the AGV indicating that charging is complete and the AGV is ready to leave. The AGV then autonomously drives out of the charging area, and the system switches to a low-power monitoring mode, waiting for the arrival of the next AGV to trigger the switch.

[0080] Finally, the following points should be noted: First, in the description of this application, it should be noted that, unless otherwise specified and limited, the terms "installation", "connection", and "linkage" should be interpreted broadly, and can be mechanical or electrical connections, or internal connections between two components, or direct connections. "Up", "down", "left", "right", etc. are only used to indicate relative positional relationships. When the absolute position of the described object changes, the relative positional relationship may change.

[0081] Secondly: The accompanying drawings of the embodiments disclosed in this invention only involve the structures involved in the embodiments disclosed in this invention. Other structures can refer to the general design. In the absence of conflict, the same embodiment and different embodiments of this invention can be combined with each other.

[0082] In conclusion, the above description is only a preferred embodiment of the present invention and is not intended to limit the present invention. 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.

[0083] The above are all preferred embodiments of this application, and are not intended to limit the scope of protection of this application. Therefore, all equivalent changes made in accordance with the structure, shape and principle of this application should be covered within the scope of protection of this application.

Claims

1. A wireless charging control and guidance system for automated guided vehicles, characterized in that: The system is initiated by the charging base station to guide the centering process, including: A charging base station includes a fixed base, a multi-dimensional motion mechanism mounted on the base, and a wireless transmission module mounted at the moving end of the motion mechanism; the motion mechanism has at least mutually orthogonal translational degrees of freedom in the horizontal plane in the X and Y directions. The binocular vision perception module is fixedly installed on a rigid bracket on the charging base station, located outside the working range of the travel motion mechanism. It is used to synchronously acquire the left and right eye images of the receiving module area at the bottom of the automated guided vehicle parked in the charging area from a fixed viewpoint, and generate a dense parallax map covering the surface area of ​​the receiving module through stereo matching. The pose calculation engine is connected to the binocular vision perception module and is used to reconstruct the three-dimensional dense point cloud of the receiving module based on the dense disparity map. The three-dimensional dense point cloud is then registered in three dimensions with the pre-stored standard geometric model of the receiving module. The engine calculates the six-degree-of-freedom pose deviation vector of the receiving module relative to the current reference position of the wireless transmitting module. The six-degree-of-freedom pose deviation vector includes X-axis linear displacement deviation, Y-axis linear displacement deviation, Z-axis linear displacement deviation, and angular displacement deviation around the X-axis, Y-axis, and Z-axis. An adaptive centering controller is connected to the pose calculation engine and the drive units of each axis of the travel motion mechanism. It is used to generate motion commands with the six-degree-of-freedom pose deviation vector as feedback input and a visual servo control strategy to drive the travel motion mechanism to move the wireless transmission module continuously in the horizontal and / or vertical directions. During the movement, the binocular vision perception module continuously updates the feedback to form a closed-loop adjustment until the comprehensive deviation between the wireless transmission module and the receiving module converges to within the preset centering tolerance band. The charging management unit is communicatively connected to the power conversion circuit of the wireless transmission module and the on-board battery management system of the automated guided vehicle. It is used to start the wireless charging process after centering and locking, and to control the travel motion mechanism to reset to the standby position after charging is completed.

2. The wireless charging control and guidance system for an automated guided vehicle according to claim 1, characterized in that: The travel motion mechanism is an XY two-dimensional precision slide or an XYZ three-dimensional precision slide. Each of its motion axes adopts ball screw transmission and precision guide rail guidance. It is driven by an independent servo motor and equipped with an absolute position feedback encoder. The single-axis repeatability is better than ±0.05mm, and the X and Y strokes are not less than ±150mm to cover the typical stopping deviation range of the automated guided vehicle.

3. The wireless charging control and guidance system for an automated guided vehicle according to claim 2, characterized in that: The binocular vision perception module includes a first global shutter industrial camera and a second global shutter industrial camera arranged at intervals along a horizontal baseline, with a baseline distance of 80mm to 200mm. The optical axes of the two cameras are installed at a fixed angle to the horizontal plane to view the receiving module area from above. Both cameras are equipped with an active infrared fill light module, and the activation and brightness of the infrared fill light module are automatically controlled by an ambient light sensor.

4. A wireless charging control and guidance system for an automated guided vehicle according to claim 3, characterized in that: The standard geometric model of the receiving module pre-stored in the pose calculation engine is a three-dimensional mesh model generated based on the CAD design file of the receiving module, or a reference point cloud model obtained by three-dimensional scanning of the entity receiving module in a standard pose. The three-dimensional registration performed by the pose calculation engine adopts a two-level registration strategy: first, coarse registration is performed based on the principal component analysis method of the point cloud normal vector distribution, and then fine registration is performed using the iterative nearest point algorithm.

5. A wireless charging control and guidance system for an automated guided vehicle according to claim 4, characterized in that: The adaptive centering controller adopts a cascaded control structure, with its outer loop being a model prediction controller or sliding mode controller based on visual pose error, and its inner loop being the current loop and speed loop of each drive shaft; the feedback update frequency of the binocular vision perception module is not less than 20Hz, and the inner loop control cycle is not less than 1kHz.

6. A wireless charging control and guidance system for an automated guided vehicle according to claim 5, characterized in that: The wireless transmitting module includes a high-frequency full-bridge inverter circuit, an LCC-type compensation network, and a transmitting coil; the receiving module includes a receiving coil, a compensation network, and a rectifier filter circuit; the operating frequency between the wireless transmitting module and the receiving module is dynamically adjustable in the range of 80kHz to 300kHz to adapt to resonance matching under different air gap distances.

7. A wireless charging control and guidance system for an automated guided vehicle according to claim 6, characterized in that: It also includes a docking wake-up sensor group, which includes at least one of a buried electromagnetic induction sensor, a side-beam photoelectric sensor, or an ultrasonic sensor, installed at or inside the charging area entrance, for detecting the arrival of the automated guided vehicle in a low-power manner and waking up the binocular vision perception module and the pose calculation engine to start working after the vehicle is detected.

8. A wireless charging control and guidance system for an automated guided vehicle according to claim 7, characterized in that: The charging management unit integrates a foreign object and living object safety detection module, which includes: The impedance spectroscopy analysis submodule is used to inject a probe signal into the transmitting coil and measure the equivalent series resistance and inductance. When the changes in the detection parameters exceed a preset threshold, it determines the presence of a conductive foreign object. The visual recognition submodule is used to call a pre-trained lightweight neural network to classify and recognize the images acquired by the binocular vision perception module, and to determine whether there are foreign objects in the gap area between the transceiver coils. The impedance spectrum analysis submodule and the visual recognition submodule output alarm signals using "OR" logic. If either submodule triggers an alarm, a protective interrupt is executed.

9. A wireless charging control and guidance method using the system described in any one of claims 1 to 8, characterized in that, Includes the following steps: S1: The docking wake-up sensor group detects that the automated guided vehicle has entered the charging area and sends a wake-up signal; S2: The binocular vision perception module responds to the wake-up signal and acquires the left and right eye images of the receiving module area; S3: The pose calculation engine performs stereo matching, 3D point cloud reconstruction and model registration on the left and right eye images to calculate the six-degree-of-freedom pose deviation vector of the receiving module relative to the wireless transmitting module. S4: The adaptive centering controller uses the six-degree-of-freedom pose deviation vector as feedback to drive the stroke motion mechanism to move the wireless transmission module toward the target centering position. During the movement, steps S2 to S4 are continuously executed to form a visual servoing closed loop. S5: When all components of the six-degree-of-freedom pose deviation vector converge to the preset threshold and remain stable for no less than the preset time, centering is determined to be locked. S6: The charging management unit initiates wireless charging, monitors charging parameters in real time, and dynamically adjusts the output. S7: After charging is complete, the charging management unit cuts off the output and controls the stroke motion mechanism to reset to the initial position; S8: The system switches to low power monitoring mode and waits for the next trigger.

10. The method according to claim 9, characterized in that, In step S3, when the pose calculation engine has multiple standard geometric models of receiving modules corresponding to different vehicle models stored in advance, coarse registration is performed on each model in sequence before 3D registration, and the model with the smallest registration residual is selected as the current matching model to achieve multi-vehicle adaptive recognition; in step S4, after the adaptive centering controller receives the six-degree-of-freedom pose deviation vector for the first time, if it detects that the deviation of a certain axis exceeds 80% of the travel range of that axis, it will reduce the deviation to within 50% of the travel range in the open-loop fast approach mode before switching to closed-loop visual servo precise approach.

Images