A double-wheel differential AGV dead reckoning system and method

Abstract

This invention discloses a dead reckoning system and method for a dual-wheel differential AGV, belonging to the field of dead reckoning technology. The system employs a data acquisition module to acquire multi-source operating data of the AGV and preprocess it to generate a standardized data stream; a mileage calculation module to generate the original mileage increments for the left and right wheels; a heading fusion module to fuse the standardized data stream and the original mileage increments through a filter to generate a fused heading angle; a motion control module to generate motor drive commands based on the fused heading angle and target commands; a memory optimization module to identify scenarios and match historical strategies, dynamically adjusting relevant calculation parameters; and a visualization module to visualize the operating data and control process and generate a trajectory report. This invention, by fusing multi-source sensor data and intelligent memory optimization, achieves adaptive parameter adjustment and scenario-based strategy reuse in dead reckoning, effectively suppressing accumulated errors and improving the motion control accuracy and robustness of the AGV in complex environments.

Description

Technical Field

[0001] This invention relates to the field of dead reckoning technology, specifically to a dead reckoning system and method for a dual-wheel differential speed AGV. Background Technology

[0002] In the existing technology, dead reckoning of dual-wheel differential AGV is mainly based on a filtering fusion scheme of wheel speed odometer and inertial measurement unit. This scheme continuously fuses wheel speed data from encoder and angular velocity data from IMU through a predetermined filter to obtain a more stable heading estimate than a single sensor. This is a widely adopted classic method, and its core technology lies in designing a fixed data fusion model. However, existing technologies have certain shortcomings. First, the key parameters of their fusion model, such as filter gain or noise statistics, are usually statically set in advance according to specific experimental environments during the system debugging phase. Once the actual operating environment changes, such as moving from a smooth surface to a rough road surface or experiencing load changes, these fixed parameters cannot achieve the optimal dynamic allocation of sensor data weights, resulting in a decrease in fusion accuracy. Furthermore, existing technologies lack the ability to learn from historical operating experience and self-optimize. Each time the system faces a similar scenario, it mechanically repeats the same calculation process and cannot memorize and reuse past successful control strategies. Its performance depends entirely on the generalization ability of the initial design, and its adaptability and intelligence level are insufficient in complex and ever-changing application scenarios. Therefore, developing a dead reckoning system for dual-wheel differential speed AGVs is of great significance. Summary of the Invention

[0003] The purpose of this invention is to provide a dead reckoning system and method for dual-wheel differential speed AGVs to address the shortcomings in the prior art.

[0004] To achieve the above objectives, the present invention provides the following technical solution: a dual-wheel differential speed AGV dead reckoning system, comprising: Data acquisition module: used to acquire multi-source operating data of AGV, and preprocess the multi-source operating data to generate a standardized data stream; Mileage calculation module: Connected to the data acquisition module, it is used to calculate the original mileage increment of the left and right wheels based on a standardized data stream; Heading fusion module: Connected to the mileage calculation module, it fuses the standardized data stream and the original mileage increment through a filter to generate a fused heading angle; Motion control module; connected to the heading fusion module, generates motor drive commands based on the fused heading angle and target motion commands; Memory optimization module: Connected to the motion control module, it is used to identify the current motion scenario and match historical optimization strategies based on standardized data streams, and dynamically adjust the parameters for mileage calculation and heading angle calculation; Visualization module: Connected to the memory optimization module, it is used to visualize the multi-source operating data and control process of the AGV and generate the motion trajectory report of the AGV.

[0005] In a preferred embodiment, the data acquisition module includes: The sensor interface unit is used to collect sensor data, which includes wheel speed pulse signals and vehicle inertial measurement data, including angular velocity and acceleration values. A communication interface unit is used to receive target motion commands. The data preprocessing unit is used to filter and perform timing correction on the wheel speed pulse signal and vehicle inertial measurement data to generate a standardized data stream.

[0006] In a preferred embodiment, the mileage calculation module includes: The wheel speed calculation unit calculates the instantaneous linear velocities of the left and right wheels based on the wheel speed pulse signals; The time interval between adjacent pulse signals and the number of pulses per revolution within a preset time period are measured. The instantaneous rotational speeds of the left and right wheels are calculated based on the time interval and the number of pulses per revolution. The instantaneous linear velocities of the left and right wheels are generated based on the preset wheel radius. The integral calculation unit integrates the instantaneous linear velocities of the left and right wheels to generate the original mileage increments of the left and right wheels.

[0007] In a preferred embodiment, the heading fusion module includes: The inertial calculation unit is used to integrate the angular velocity value to obtain the inertial heading angle increment; The odometer calculation unit calculates the odometer heading angle increment based on the original odometer increments of the left and right wheels. The filtering and fusion unit is used to fuse the inertial heading angle increment and the odometer heading angle increment to generate a fused heading angle.

[0008] In a preferred embodiment, the motion control module includes: The dual-loop control unit includes a speed closed-loop control subunit and a heading closed-loop control subunit; The speed closed-loop control subunit controls the left and right wheel speeds of the AGV through two PID controllers, and generates the target heading based on the adjustment of the left and right wheel speeds. The heading closed-loop control subunit uses the fused heading angle as feedback, compares it with the target heading, and outputs the differential compensation amount; The instruction synthesis unit is used to superimpose the differential compensation amount onto the target speeds of the left and right wheels to generate motor drive instructions.

[0009] In a preferred embodiment, the memory optimization module includes: The scene feature extraction unit is used to extract real-time scene feature vectors from the standardized data stream. The real-time scene feature vectors include environmental feature components that characterize the external operating environment and operating state feature components that characterize the instantaneous motion state of the AGV itself. The historical strategy unit is used to store multiple historical strategy records. Each historical strategy record contains a historical scene feature template and an associated set of historical valid control parameters. The scene matching unit is used to calculate the comprehensive matching degree between the real-time scene feature vector and each historical scene feature template; The parameter preloading unit is used to preload the set of historical effective control parameters from the corresponding historical strategy record into the motion control module when the overall matching degree exceeds the preset matching threshold, and start an experimental running cycle of preset duration. The experimental error between the fused heading angle and the target heading, calculated based on preload parameters, was monitored in real time during the experimental operation cycle. When the experimental error exceeds the preset safety threshold, the experimental cycle is immediately terminated, and the motion control module is triggered to recalculate the differential compensation amount.

[0010] In a preferred embodiment, the step of calculating the comprehensive matching degree between the real-time scene feature vector and each historical scene feature template is as follows: The similarity between the environmental feature component and the running state feature component in the real-time scene feature vector and the corresponding component in the corresponding historical scene feature template is calculated to obtain the environmental similarity and state similarity. Dynamic weights are assigned to environmental similarity and state similarity based on the current AGV control mode. The control modes include uniform linear motion mode, acceleration, deceleration, and turning mode; In uniform linear motion mode, increase the weight of environmental similarity; in acceleration, deceleration and turning modes, increase the weight of state similarity. The weighted environmental similarity and state similarity are combined to generate the comprehensive matching degree.

[0011] In a preferred embodiment, the visualization module includes: The trajectory drawing unit, based on the original mileage increment and the fused heading angle, draws and displays the calculated motion trajectory of the AGV in real time; The feature visualization unit is used to receive and display real-time scene feature vectors, matched historical scene feature templates, and corresponding comprehensive matching degrees. The control status visualization unit is used to display the differential compensation amount, the status of the experimental cycle, and the comparison between the experimental error and the preset safety threshold in real time. The outputs of the trajectory drawing unit, feature visualization unit, and control status visualization unit are used as the motion trajectory report of the AGV.

[0012] This invention also provides a method for dead reckoning using a dual-wheel differential AGV, comprising: Acquire multi-source operation data of AGVs, preprocess the multi-source operation data, and generate a standardized data stream; The original mileage increments for the left and right wheels are calculated based on standardized data streams; The fused heading angle is generated by fusing the standardized data stream and the original mileage increment through a filter. Based on the fusion of heading angle and target motion commands, motor drive commands are generated; Based on standardized data streams, the current motion scenario is identified and historical optimization strategies are matched to dynamically adjust the parameters for mileage calculation and heading angle calculation. The multi-source operation data and control process of AGVs are visualized to generate AGV motion trajectory reports.

[0013] The technical effects and advantages provided by the present invention in the above technical solution are as follows: 1. This invention designs a dual-closed-loop motion control module based on fused heading feedback, which to some extent solves the inherent deviation problem of dual-wheel differential AGVs caused by slight differences in motor characteristics, tire wear, or ground conditions. The system does not simply perform data fusion, but uses the fused high-precision heading angle as the core feedback quantity to construct an external heading closed-loop controller. This controller continuously compares the current fused heading with the target heading, calculates the accurate differential compensation amount in real time, and dynamically applies this compensation amount to the target speed of the left and right wheels in an addition and subtraction manner. Furthermore, the left and right wheels are finely controlled through dual controllers. This active and continuous closed-loop compensation mechanism enables the AGV to be guided as if by an invisible guide rail, effectively resisting various internal asymmetries and external interferences, achieving stable and accurate straight-line travel and fixed-point turning, and maximizing the performance potential of low-cost hardware. 2. This invention introduces a memory optimization module, enabling the system to possess an intelligent optimization mechanism for scene recognition and strategy self-optimization. This module can extract a hybrid feature vector reflecting the external environment and its own motion state in real time and quickly match it with a historical strategy library. When a match is successful, the system can proactively preload historically validated effective control parameters, realizing the reuse of control experience and significantly improving response speed and control accuracy in known or similar scenarios. Simultaneously, an experimental preloading and safety verification mechanism is employed. After introducing historical parameters, a monitored experimental run cycle is initiated. If performance fails to meet expectations, the system automatically reverts and recalculates, ensuring absolute safety and robustness in the optimization process. Based on this mechanism, the AGV transforms from a passively responding control machine into an intelligent agent capable of accumulating experience and adapting to the environment. Attached Figure Description

[0014] To more clearly illustrate the technical solutions in the embodiments of this application or the prior art, the drawings used in the embodiments will be briefly introduced below. Obviously, the drawings described below are only some embodiments recorded in this invention. For those skilled in the art, other drawings can be obtained based on these drawings.

[0015] Figure 1 This is a system flowchart of the present invention.

[0016] Figure 2 This is a flowchart of the method of the present invention.

[0017] Figure 3 This is a logic block diagram of the present invention. Detailed Implementation

[0018] To make the objectives, technical solutions, and advantages of the embodiments of the present invention clearer, 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, 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.

[0019] Example 1, please refer to Figure 1 and Figure 3 As shown in this embodiment, a dual-wheel differential AGV dead reckoning system includes: Data acquisition module: used to acquire multi-source operating data of AGV, and preprocess the multi-source operating data to generate a standardized data stream; Mileage calculation module: Connected to the data acquisition module, it is used to calculate the original mileage increment of the left and right wheels based on a standardized data stream; Heading fusion module: Connected to the mileage calculation module, it fuses the standardized data stream and the original mileage increment through a filter to generate a fused heading angle; Motion control module; connected to the heading fusion module, generates motor drive commands based on the fused heading angle and target motion commands; Memory optimization module: Connected to the motion control module, it is used to identify the current motion scenario and match historical optimization strategies based on standardized data streams, and dynamically adjust the parameters for mileage calculation and heading angle calculation; Visualization module: Connected to the memory optimization module, it is used to visualize the multi-source operating data and control process of the AGV and generate the motion trajectory report of the AGV; Furthermore, in the existing technology, dead reckoning of dual-wheel differential AGV is mainly based on a filtering fusion scheme of wheel speed odometer and inertial measurement unit. This scheme continuously fuses wheel speed data from encoder and angular velocity data from IMU through a predetermined filter to obtain a more stable heading estimate than a single sensor. This is a widely adopted classic method, and its core technology lies in designing a fixed data fusion model. However, existing technologies have certain shortcomings. First, the key parameters of their fusion model, such as filter gain or noise statistics, are usually statically set in advance according to specific experimental environments during the system debugging phase. Once the actual operating environment changes, such as moving from a smooth surface to a rough road surface or experiencing load changes, these fixed parameters cannot achieve the optimal dynamic allocation of sensor data weights, resulting in a decrease in fusion accuracy. Furthermore, existing technologies lack the ability to learn from historical operating experience and self-optimize. Each time the system faces a similar scenario, it mechanically repeats the same calculation process and cannot memorize and reuse past successful control strategies. Its performance depends entirely on the generalization ability of the initial design, and its adaptability and intelligence level are insufficient in complex and ever-changing application scenarios. This invention addresses the inherent deviation problem of dual-wheel differential AGVs caused by slight differences in motor characteristics, tire wear, or ground conditions by designing a dual-closed-loop motion control module based on fused heading feedback. The system does not simply fuse data; instead, it uses the fused high-precision heading angle as the core feedback quantity to construct an external heading closed-loop controller. This controller continuously compares the current fused heading with the target heading, calculates the precise differential compensation in real time, and dynamically applies this compensation to the target speeds of the left and right wheels in an additive and subtractive manner. Furthermore, the dual controllers provide fine control over the left and right wheels. This proactive and continuous closed-loop compensation mechanism allows the AGV to be guided as if by an invisible guide rail, effectively counteracting various internal asymmetries and external interferences, achieving stable and precise straight-line movement and point-to-point turning, maximizing the performance potential of low-cost hardware. By introducing a memory optimization module, the system acquires an intelligent optimization mechanism for scene recognition and strategy self-optimization. This module can extract a hybrid feature vector reflecting the external environment and its own motion state in real time and quickly match it with a historical strategy library. When a match is successful, the system can proactively preload historically validated effective control parameters, enabling the reuse of control experience and significantly improving response speed and control accuracy in known or similar scenarios. Simultaneously, an experimental preloading and safety verification mechanism is employed. After introducing historical parameters, a monitored experimental run cycle is initiated. If performance fails to meet expectations, the system automatically rolls back and recalculates, ensuring absolute safety and robustness in the optimization process. Based on this mechanism, the AGV transforms from a passively responding control machine into an intelligent agent capable of accumulating experience and adapting to the environment.

[0020] In one embodiment, the data acquisition module includes: The sensor interface unit is used to collect sensor data, which includes wheel speed pulse signals and vehicle inertial measurement data, including angular velocity and acceleration values. A communication interface unit is used to receive target motion commands. The data preprocessing unit is used to filter and perform timing correction on the wheel speed pulse signal and vehicle body inertial measurement data to generate a standardized data stream; Furthermore, the sensor interface unit directly captures the wheel speed pulse signal from the Hall encoder through the microcontroller's general-purpose input / output pins. It utilizes the microcontroller's input capture function to accurately measure the time interval between the rising and falling edges of the pulse. Simultaneously, it continuously reads the raw register data of the angular velocity and acceleration values ​​from the inertial measurement unit chip via the integrated circuit bus interface at a preset sampling rate. The communication interface unit receives target motion commands from the host computer via a universal asynchronous transceiver serial port at a specific baud rate. The command format is a fixed-length data frame containing a start character, command word, velocity parameters, checksum, and end character. After reception... Frame check sequence verification is performed to ensure instruction integrity. The data preprocessing unit first performs software de-jitter filtering on the wheel speed pulse signal, and uses the continuous sampling majority decision method to eliminate erroneous pulses caused by mechanical jitter of the contact points. Then, the inertial measurement data is filtered by sliding window mean to suppress high-frequency noise. Next, all sensor data are marked with a unified time stamp using the microcontroller's internal high-precision timer, and the data streams of each sensor are aligned to the same time base through an interpolation algorithm. Finally, the processed time-stamped data is encapsulated in a predetermined format into a standardized data stream containing a data header, timestamp, sensor identifier, data payload, and cyclic redundancy check code.

[0021] In one embodiment, the mileage calculation module includes: The wheel speed calculation unit calculates the instantaneous linear velocities of the left and right wheels based on the wheel speed pulse signals; The time interval between adjacent pulse signals and the number of pulses per revolution within a preset time period are measured. The instantaneous rotational speeds of the left and right wheels are calculated based on the time interval and the number of pulses per revolution. The instantaneous linear velocities of the left and right wheels are generated based on the preset wheel radius. The integral calculation unit integrates the instantaneous linear velocities of the left and right wheels to generate the original mileage increments of the left and right wheels; Furthermore, the wheel speed calculation unit first acquires the wheel speed pulse signals of the left and right wheels respectively through the input capture function of the microcontroller. Specifically, when the rising edge of the pulse is captured, the count value of the high-precision timer is recorded. The time interval between adjacent pulse signals is obtained by calculating the timer count difference between two consecutive rising edges. At the same time, based on the single-cycle pulse number parameter pre-stored in the flash memory, the reciprocal of the time interval is converted into the instantaneous rotation speed of the corresponding wheel. Then, the instantaneous rotation speed is multiplied by the pre-stored wheel radius parameter and then by twice the constant of pi, thereby generating the instantaneous linear velocity of the left wheel and the instantaneous linear velocity of the right wheel respectively. The integral calculation unit uses a fixed control cycle as the time base. Within each control cycle, it multiplies the latest calculated instantaneous linear velocity of the left wheel and the instantaneous linear velocity of the right wheel by the duration of the control cycle to obtain the minute displacement increment of the left and right wheels within that cycle. It then accumulates these increments into the registers for the total original mileage increment of the left wheel and the right wheel, respectively, using an integral accumulation method. An overflow management mechanism is also introduced, which automatically converts the accumulated value to a higher precision floating-point number or a long integer for storage when it exceeds the register length, thereby continuously generating high-precision original mileage increments for the left and right wheels.

[0022] In one embodiment, the heading fusion module includes: The inertial calculation unit is used to integrate the angular velocity value to obtain the inertial heading angle increment; The odometer calculation unit calculates the odometer heading angle increment based on the original odometer increments of the left and right wheels. A filtering and fusion unit is used to fuse the inertial heading angle increment and the odometer heading angle increment to generate a fused heading angle; Furthermore, the inertial calculation unit reads the angular velocity values ​​of the inertial measurement unit from the standardized data stream at fixed sampling periods, and multiplies the angular velocity values ​​acquired in each sampling period by the sampling period duration to obtain the minute angle change within that minute time interval. Then, an accumulation integration algorithm is used to sum the minute angle changes across all consecutive time intervals. Simultaneously, an automatic zero-bias calibration algorithm is introduced to eliminate the static drift error of the gyroscope, thus obtaining a continuous and relatively accurate inertial heading angle increment. The odometer calculation unit, based on the original odometer increments of the left and right wheels provided by the odometer calculation module, first calculates the difference between the original odometer increments of the right and left wheels within the same time interval. Then, this difference is divided by the wheel spacing constant preset in the system parameters, and the odometer heading angle increment derived purely from wheel speed within that time interval is directly calculated based on the dual-wheel differential kinematic model. The filtering and fusion unit uses a complementary filter designed based on frequency response characteristics to fuse the high-frequency inertial heading angle increment with the low-frequency stable odometer heading angle increment. The specific implementation employs a complementary filter structure, the core of which lies in dynamically configuring the cutoff frequencies of the high-pass and low-pass filters to achieve optimal fusion. This unit first analyzes in real-time the high-frequency noise power spectral density of the inertial measurement unit's angular velocity signal and the confidence index of the odometer heading angle increment signal. The confidence index assesses whether slippage has occurred by calculating the correlation between the left and right wheel speeds within a continuous time window. Then, based on the quantization results of the noise power spectral density and the confidence index, the real-time cutoff frequencies of the high-speed and low-pass filters are dynamically calculated and set through a preset mapping table. This automatically reduces the cutoff frequency of the high-speed filter to enhance the filtering effect when the angular velocity data noise is high, and automatically reduces the cutoff frequency of the low-pass filter to reduce its contribution when the odometer data confidence is low. Finally, the inertial heading angle increment signal processed by the dynamically tuned high-speed filter and the odometer heading angle increment signal processed by the dynamically tuned low-pass filter are time-domain superimposed, and the superposition result is normalized to generate an adaptive fused heading angle output.

[0023] In one embodiment, the motion control module includes: The dual-loop control unit includes a speed closed-loop control subunit and a heading closed-loop control subunit; The speed closed-loop control subunit controls the left and right wheel speeds of the AGV through two PID controllers, and generates the target heading based on the adjustment of the left and right wheel speeds. The heading closed-loop control subunit uses the fused heading angle as feedback, compares it with the target heading, and outputs the differential compensation amount; The instruction synthesis unit is used to superimpose the differential compensation amount onto the target speeds of the left and right wheels to generate motor drive instructions. Furthermore, the speed closed-loop control subunit uses two independent position-type PID controllers to adjust the deviations between the actual speed of the left wheel and the target speed of the left wheel, and between the actual speed of the right wheel and the target speed of the right wheel, respectively. Each PID controller includes calculations of proportional, integral, and derivative terms. The integral term uses an anti-saturation algorithm to limit accumulated error, and the derivative term uses incomplete differentiation to suppress measurement noise. The final outputs are the PWM duty cycle control values ​​for the left and right wheel motors. Simultaneously, based on the set values ​​of the target speeds of the left and right wheels and combined with the vehicle kinematics model, the desired target heading angle is calculated in real time. The heading closed-loop control subunit receives the fused heading angle from the heading fusion module as a process variable, compares it with the target heading angle provided by the speed closed-loop control subunit to obtain the heading angle deviation, which is input to a... The proportional-integral controller performs calculations, and the controller output is the dynamic differential compensation amount. The sign of this compensation amount determines the compensation direction, and its magnitude is related to the absolute value and trend of the heading angle deviation. The instruction synthesis unit superimposes the differential compensation amount output by the heading closed-loop control subunit onto the original left wheel target speed and right wheel target speed obtained from the upper-level instruction parsing, respectively, according to the principle of opposite signs and equal magnitudes. That is, the final target speed of the left wheel is equal to the original left wheel target speed minus the differential compensation amount, and the final target speed of the right wheel is equal to the original right wheel target speed plus the differential compensation amount. Then, the synthesized final target speeds of the left and right wheels are subjected to safety limiting processing to ensure that they do not exceed the maximum speed range allowed by the motor. Finally, based on the calibration mapping relationship between speed and PWM duty cycle, the left wheel motor drive command and the right wheel motor drive command that can directly drive the motor are generated.

[0024] In one embodiment, the memory optimization module includes: The scene feature extraction unit is used to extract real-time scene feature vectors from the standardized data stream. The real-time scene feature vectors include environmental feature components that characterize the external operating environment and operating state feature components that characterize the instantaneous motion state of the AGV itself. The historical strategy unit is used to store multiple historical strategy records. Each historical strategy record contains a historical scene feature template and an associated set of historical valid control parameters. The scene matching unit is used to calculate the comprehensive matching degree between the real-time scene feature vector and each historical scene feature template; The parameter preloading unit is used to preload the set of historical effective control parameters from the corresponding historical strategy record into the motion control module when the overall matching degree exceeds the preset matching threshold, and start an experimental running cycle of preset duration. The experimental error between the fused heading angle and the target heading, calculated based on preload parameters, was monitored in real time during the experimental operation cycle. When the experimental error exceeds the preset safety threshold, the experimental cycle is immediately terminated, and the motion control module is triggered to recalculate the differential compensation amount. Furthermore, the scene feature extraction unit in the memory optimization module is implemented by performing sliding window analysis on the standardized data stream, calculating the root mean square value of the high-frequency acceleration signal of the inertial measurement unit within the window as a ground roughness estimate to constitute environmental feature components, and simultaneously calculating the variance of the ratio of the pulse frequencies of the left and right wheels within the same window and the first-order difference standard deviation of the fused heading angle as operating state feature components. These components are then normalized and concatenated into a real-time scene feature vector. The history strategy unit maintains a circular queue structure in non-volatile memory using a least recently used elimination strategy to store historical strategy records. Each record, in addition to the feature template and parameter set, also includes a timestamp and the number of successful calls for record validity evaluation. The scene matching unit uses an approximate nearest neighbor search algorithm based on locality-sensitive hashing. First, a hash signature is generated for the real-time scene feature vector, quickly retrieving a subset of candidate historical templates. Then, the Euclidean distance is accurately calculated within this subset, and the environmental feature components in the distance calculation are dynamically adjusted according to whether the current control mode is uniform linear motion. The weight ratio of the operating state characteristic components is used to map the inverse of the distance to the comprehensive matching degree through the sigmoid function. After the comprehensive matching degree exceeds the set threshold, the parameter preloading unit will temporarily replace the PID parameters of the heading closed-loop control subunit in the motion control module and the wheel diameter calibration parameters in the mileage calculation module with the set of historical effective control parameters stored in the historical strategy record. At the same time, a fixed-duration experimental running cycle controlled by the system timer is started. During the experimental running cycle, the system samples the deviation between the fused heading angle and the target heading at a frequency higher than that of conventional control, calculates the moving average of its absolute value as the experimental error, and compares the experimental error with the independently set preset safety threshold in real time. Once the experimental error exceeds the preset safety threshold for three consecutive sampling points, the parameter preloading unit immediately sends an interrupt signal to the motion control module to stop the current experimental running cycle and restore all parameters to their default values. At the same time, the heading closed-loop control subunit of the motion control module is triggered to recalculate the differential compensation amount based on the default parameters, thereby ensuring system safety.

[0025] In one embodiment, the step of calculating the comprehensive matching degree between the real-time scene feature vector and each historical scene feature template is as follows: The similarity between the environmental feature component and the running state feature component in the real-time scene feature vector and the corresponding component in the corresponding historical scene feature template is calculated to obtain the environmental similarity and state similarity. Dynamic weights are assigned to environmental similarity and state similarity based on the current AGV control mode. The control modes include uniform linear motion mode, acceleration, deceleration, and turning mode; In uniform linear motion mode, increase the weight of environmental similarity; in acceleration, deceleration and turning modes, increase the weight of state similarity. The weighted environmental similarity and state similarity are combined to generate the comprehensive matching degree; Furthermore, the similarity between the environmental feature components and the operational state feature components is calculated separately. For the environmental feature components, the environmental similarity is obtained by calculating the normalized correlation coefficient between the estimated ground roughness sequence in the real-time feature vector and the corresponding sequence in the historical template. For the operational state feature components, the state similarity is obtained by calculating the inverse proportional function between the variance of wheel speed ratio and the standard deviation of heading angle change in the real-time feature vector and the corresponding values ​​in the historical template. Then, dynamic weights are assigned to these two similarities according to the current AGV control mode. The control mode is derived from parsing the target motion command. If the angular velocity command in the command is zero and the speed... If the command is constant, it is determined to be a uniform linear motion mode. In this case, the weighting coefficient of environmental similarity is set to 0.7 and the weighting coefficient of state similarity is set to 0.3. If a change in speed command or angular velocity command is detected, it is determined to be an acceleration, deceleration, or turning mode. In this case, the weighting coefficient of environmental similarity is set to 0.3 and the weighting coefficient of state similarity is set to 0.7. Finally, the weighted environmental similarity and the weighted state similarity are directly added to obtain the initial comprehensive matching degree. This initial value is then multiplied by the average historical call success rate pre-stored in the corresponding historical strategy record for secondary correction, thereby generating the final comprehensive matching degree value used for decision-making.

[0026] In one embodiment, the visualization module includes: The trajectory drawing unit, based on the original mileage increment and the fused heading angle, draws and displays the calculated motion trajectory of the AGV in real time; The feature visualization unit is used to receive and display real-time scene feature vectors, matched historical scene feature templates, and corresponding comprehensive matching degrees. The control status visualization unit is used to display the differential compensation amount, the status of the experimental cycle, and the comparison between the experimental error and the preset safety threshold in real time. The outputs of the trajectory drawing unit, feature visualization unit, and control status visualization unit are used as the motion trajectory report of the AGV. Furthermore, the trajectory drawing unit in the visualization module is implemented by establishing a global coordinate system, combining the original mileage increments of the left and right wheels provided by the mileage calculation module with the fused heading angle output by the heading fusion module, and using the dead reckoning algorithm to calculate the current position coordinates of the AGV in real time. The coordinate points under continuous timestamps are stored in a circular buffer and dynamically drawn on the grid map layer of the display interface in the form of polylines with specific colors and line widths through the graphical interface, realizing real-time visualization of the calculated motion trajectory. The feature visualization unit adopts a multi-view linkage method, drawing the real-time scene feature vector output by the scene feature extraction unit in the form of a radar chart in the first view area, where each axis represents a different environmental or state feature dimension, and the values ​​are normalized. The matching historical scene feature templates retrieved by the scene matching unit are drawn in the form of a comparison bar chart in the second view area. The calculated comprehensive matching degree is displayed in the third view area as a dynamic progress bar and numerical percentage, and the matching degree is reflected by color gradient. The control status visualization unit subscribes to the differential compensation data released by the motion control module in real time, plots it as a curve that fluctuates over time, and displays it in the fourth view area. At the same time, it uses eye-catching icons and color changes to indicate the activation, progress, or termination status of the experimental cycle. It also plots the real-time collected experimental error and the preset safety threshold as a hyperbola in the same view area, and uses the relative position and fill color to intuitively reflect whether the error exceeds the limit. Finally, the module automatically takes a snapshot of the graphical data and key values ​​of all the above views at preset time intervals, adds timestamps and annotations, and generates a structured AGV motion trajectory report document.

[0027] Example 2, please refer to Figure 2 As shown in this embodiment, a dead reckoning method for a dual-wheel differential AGV includes: Acquire multi-source operation data of AGVs, preprocess the multi-source operation data, and generate a standardized data stream; The original mileage increments for the left and right wheels are calculated based on standardized data streams; The fused heading angle is generated by fusing the standardized data stream and the original mileage increment through a filter. Based on the fusion of heading angle and target motion commands, motor drive commands are generated; Based on standardized data streams, the current motion scenario is identified and historical optimization strategies are matched to dynamically adjust the parameters for mileage calculation and heading angle calculation. The multi-source operation data and control process of AGVs are visualized to generate AGV motion trajectory reports.

[0028] The above description is merely a specific embodiment of this application, but the scope of protection of this application is not limited thereto. Any variations or substitutions that can be easily conceived by those skilled in the art within the scope of the technology disclosed in this application should be included within the scope of protection of this application. Therefore, the scope of protection of this application should be determined by the scope of the claims.

Claims

1. A dead reckoning system for a dual-wheel differential speed AGV, characterized in that, Data acquisition module: used to acquire multi-source operating data of AGV, and preprocess the multi-source operating data to generate a standardized data stream; Mileage calculation module: Connected to the data acquisition module, it is used to calculate the original mileage increment of the left and right wheels based on a standardized data stream; Heading fusion module: Connected to the mileage calculation module, it fuses the standardized data stream and the original mileage increment through a filter to generate a fused heading angle; Motion control module; connected to the heading fusion module, generates motor drive commands based on the fused heading angle and target motion commands; Memory optimization module: Connected to the motion control module, it is used to identify the current motion scenario and match historical optimization strategies based on standardized data streams, and dynamically adjust the parameters for mileage calculation and heading angle calculation; Visualization module: Connected to the memory optimization module, it is used to visualize the multi-source operating data and control process of the AGV and generate the motion trajectory report of the AGV.

2. The dual-wheel differential AGV dead reckoning system according to claim 1, wherein, The data acquisition module includes: The sensor interface unit is used to collect sensor data, which includes wheel speed pulse signals and vehicle inertial measurement data, including angular velocity and acceleration values. A communication interface unit is used to receive target motion commands. The data preprocessing unit is used to filter and perform timing correction on the wheel speed pulse signal and vehicle inertial measurement data to generate a standardized data stream.

3. The double-wheel differential AGV dead reckoning system according to claim 1, wherein, The mileage calculation module includes: The wheel speed calculation unit calculates the instantaneous linear velocities of the left and right wheels based on the wheel speed pulse signals; The time interval between adjacent pulse signals and the number of pulses per revolution within a preset time period are measured. The instantaneous rotational speeds of the left and right wheels are calculated based on the time interval and the number of pulses per revolution. The instantaneous linear velocities of the left and right wheels are generated based on the preset wheel radius. The integral calculation unit integrates the instantaneous linear velocities of the left and right wheels to generate the original mileage increments of the left and right wheels.

4. The dual-wheel differential AGV dead reckoning system of claim 1, wherein, The heading fusion module includes: The inertial calculation unit is used to integrate the angular velocity value to obtain the inertial heading angle increment; The odometer calculation unit calculates the odometer heading angle increment based on the original odometer increments of the left and right wheels. The filtering and fusion unit is used to fuse the inertial heading angle increment and the odometer heading angle increment to generate a fused heading angle.

5. The dual-wheel differential AGV dead reckoning system of claim 1, wherein, The motion control module includes: The dual-loop control unit includes a speed closed-loop control subunit and a heading closed-loop control subunit; The speed closed-loop control subunit controls the left and right wheel speeds of the AGV through two PID controllers, and generates the target heading based on the adjustment of the left and right wheel speeds. The heading closed-loop control subunit uses the fused heading angle as feedback, compares it with the target heading, and outputs the differential compensation amount; The instruction synthesis unit is used to superimpose the differential compensation amount onto the target speeds of the left and right wheels to generate motor drive instructions.

6. The dual-wheel differential AGV dead reckoning system of claim 1, wherein, The memory optimization module includes: The scene feature extraction unit is used to extract real-time scene feature vectors from the standardized data stream. The real-time scene feature vectors include environmental feature components that characterize the external operating environment and operating state feature components that characterize the instantaneous motion state of the AGV itself. The historical strategy unit is used to store multiple historical strategy records. Each historical strategy record contains a historical scene feature template and an associated set of historical valid control parameters. The scene matching unit is used to calculate the comprehensive matching degree between the real-time scene feature vector and each historical scene feature template; The parameter preloading unit is used to preload the set of historical effective control parameters from the corresponding historical strategy record into the motion control module when the overall matching degree exceeds the preset matching threshold, and start an experimental running cycle of preset duration. The experimental error between the fused heading angle and the target heading, calculated based on preload parameters, was monitored in real time during the experimental operation cycle. When the experimental error exceeds the preset safety threshold, the experimental cycle is immediately terminated, and the motion control module is triggered to recalculate the differential compensation amount.

7. The dual-wheel differential AGV dead reckoning system of claim 6, wherein, The steps for calculating the comprehensive matching degree between the real-time scene feature vector and each historical scene feature template are as follows: The similarity between the environmental feature component and the running state feature component in the real-time scene feature vector and the corresponding component in the corresponding historical scene feature template is calculated to obtain the environmental similarity and state similarity. Dynamic weights are assigned to environmental similarity and state similarity based on the current AGV control mode. The control modes include uniform linear motion mode, acceleration, deceleration, and turning mode; In uniform linear motion mode, increase the weight of environmental similarity; in acceleration, deceleration and turning modes, increase the weight of state similarity. The weighted environmental similarity and state similarity are combined to generate the comprehensive matching degree.

8. The dual-wheel differential AGV dead reckoning system of claim 1, wherein, The visualization module includes: The trajectory drawing unit, based on the original mileage increment and the fused heading angle, draws and displays the calculated motion trajectory of the AGV in real time; The feature visualization unit is used to receive and display real-time scene feature vectors, matched historical scene feature templates, and corresponding comprehensive matching degrees. The control status visualization unit is used to display the differential compensation amount, the status of the experimental cycle, and the comparison between the experimental error and the preset safety threshold in real time. The outputs of the trajectory drawing unit, feature visualization unit, and control status visualization unit are used as the motion trajectory report of the AGV.

9. A method for dead reckoning of a dual-wheel differential AGV, used to implement the dual-wheel differential AGV dead reckoning system described in any one of claims 1-8, characterized in that, Acquire multi-source operation data of AGVs, preprocess the multi-source operation data, and generate a standardized data stream; The original mileage increments for the left and right wheels are calculated based on standardized data streams; The fused heading angle is generated by fusing the standardized data stream and the original mileage increment through a filter. Based on the fusion of heading angle and target motion commands, motor drive commands are generated; Based on standardized data streams, the current motion scenario is identified and historical optimization strategies are matched to dynamically adjust the parameters for mileage calculation and heading angle calculation. The multi-source operation data and control process of AGVs are visualized to generate AGV motion trajectory reports.

Images