An amr inter-factory logistics intelligent scheduling system based on composite navigation

Abstract

The application relates to the technical field of intelligent logistics, and discloses an AMR plant logistics intelligent scheduling system based on composite navigation, which comprises an information management layer, a control scheduling layer and an on-site execution layer. A vehicle-mounted controller module uses the kinematics data of an inertial measurement unit to perform time alignment and weighted fusion on the heterogeneous data of a laser radar and a bottom vision camera, solves real-time pose, and combines a navigation confidence change trend, network delay and load quality data to calculate a request safety envelope radius based on a multi-dimensional state coupling model. A robot scheduling module receives vehicle data, executes an asymmetric negotiation strategy when detecting that the request safety envelope has spatial overlap, determines the target safety envelope size of each party in conflict according to a comprehensive capability score, keeps the envelope of the disadvantaged party unchanged and forcibly compresses the envelope of the advantaged party to a limit safety radius. The application solves the safety and efficiency problems of multi-vehicle collaborative operation under complex working conditions through multi-dimensional state coupling and an asymmetric scheduling mechanism.

Description

Technical Field

[0001] This invention relates to the field of intelligent logistics technology, specifically to an AMR workshop logistics intelligent scheduling system based on composite navigation. Background Technology

[0002] With the improvement of industrial automation, autonomous mobile robots (AMRs) have been widely used in material handling in workshop logistics. In complex workshop environments, to ensure the stability and adaptability of positioning, existing AMR systems typically employ a composite navigation technology that combines LiDAR SLAM with ground QR code visual recognition. Simultaneously, a central scheduling system deployed on a server performs path planning and traffic control for multiple AMRs to achieve automated production logistics.

[0003] However, existing AMR scheduling and control technologies still have limitations in real-world high-load and high-dynamic operating scenarios. At the single-machine control level, current obstacle avoidance logic mainly relies on preset fixed safety zones or linear distance adjustments based solely on the current driving speed. This approach fails to fully consider the coupled impact of communication link quality, physical load variations, and navigation positioning accuracy on actual braking performance. For example, when fluctuations in the workshop's wireless network increase communication latency, or when a vehicle is fully loaded, leading to increased mechanical braking inertia, the theoretical safety zone generated solely based on speed is often smaller than the actual required physical braking distance, thus posing a safety hazard. Conversely, setting an excessively large fixed safety margin to cover all extreme conditions can limit the vehicle's ability to pass through narrow passages, reducing workshop space utilization. Furthermore, if heterogeneous sensors such as LiDAR and vision cameras lack strict time reference alignment, data asynchrony during high-speed vehicle movement can easily lead to jitter and divergence in positioning calculations.

[0004] At the multi-vehicle collaborative scheduling level, existing traffic control strategies typically employ relatively rigid rules, such as "first-come, first-served" zone locking mechanisms or stop-and-wait strategies based on fixed priorities. This traditional scheduling model lacks dynamic evaluation of vehicles' real-time control capabilities, treating vehicles in different states as standard individuals with equal maneuverability. When vehicles with deteriorating perception environments or delayed network responses encounter vehicles in good condition in confined spaces, the system often fails to flexibly allocate passage space based on the differences in their real-time capabilities. This can easily lead to situations where a single vehicle's poor condition causes all vehicles at the intersection to stop and wait, resulting in a loss of overall logistics efficiency.

[0005] Therefore, this invention proposes an AMR workshop logistics intelligent scheduling system based on composite navigation to address the shortcomings of existing technologies. Summary of the Invention

[0006] To address the shortcomings of existing technologies, this invention provides an intelligent scheduling system for AMR workshop logistics based on composite navigation. This system solves the problems of insufficient dynamic obstacle avoidance safety in existing AMR systems under complex working conditions due to the lack of multi-dimensional state coupling, and low efficiency of multi-machine collaborative operation due to rigid scheduling strategies.

[0007] To achieve the above objectives, the present invention provides the following technical solution: an AMR workshop logistics intelligent scheduling system based on composite navigation, comprising an information management layer, a control scheduling layer, and an on-site execution layer;

[0008] The information management layer includes a manufacturing execution module; the control and scheduling layer includes a robot scheduling module and a communication network module, the communication network module including a network clock server that provides a unified clock reference for the entire system; the field execution layer includes an autonomous mobile robot, the autonomous mobile robot being equipped with a composite navigation module, a wireless communication module, a drive module, and an on-board controller module; the composite navigation module includes a lidar, a bottom vision camera, an inertial measurement unit, and a load detection sensor.

[0009] The vehicle controller module is configured to use the time reference of the network clock server to perform time alignment and weighted fusion of heterogeneous sensor data from the LiDAR and bottom vision camera using the kinematic data of the inertial measurement unit, and calculate the real-time pose and navigation confidence.

[0010] The vehicle controller module is also configured to calculate the requested safety envelope radius based on a multi-dimensional state coupling model according to the robot's current driving speed, the trend of navigation confidence, the network latency monitored by the wireless communication module, and the load quality data collected by the load detection sensor, and upload the data packet containing the requested safety envelope radius and the vehicle state vector to the robot scheduling module.

[0011] The robot scheduling module is configured to detect whether there is spatial overlap in the requested safety envelopes of each of the host mobile robots, and when spatial overlap is detected, calculate the comprehensive capability score of each conflicting party based on the vehicle state vector, execute an asymmetric negotiation strategy to determine the target safety envelope size of each conflicting party and issue envelope control commands.

[0012] The onboard controller module is also configured to adjust the safety envelope size in response to envelope control commands and drive the autonomous mobile robot to perform actions.

[0013] Preferably, the vehicle controller module is specifically configured to: synchronize the system clock with the network clock server via the wireless communication module, and establish a sliding data window in memory to cache the original sensor data with global timestamps; determine the master clock reference time, and retrieve historical sampling data frames of asynchronous sensors from the sliding data window; perform integral calculations using the linear acceleration and angular velocity output by the inertial measurement unit to calculate the pose increment from the historical sampling time to the master clock reference time; compensate the pose increment into the original pose data corresponding to the historical sampling data frame to obtain the aligned pose, and perform weighted fusion of the aligned poses of each sensor according to the current navigation area attributes.

[0014] Preferably, the vehicle controller module is specifically configured to calculate the navigation confidence trend by: calculating the instantaneous navigation confidence based on the matching residual of sensor measurement data, and smoothing the instantaneous navigation confidence using a low-pass filter to generate a smoothed navigation confidence; performing differential operations on the smoothed navigation confidence in the time domain to calculate the smoothed confidence gradient; monitoring the direction and magnitude of the smoothed confidence gradient in real time, and determining that the positioning quality is deteriorating when the smoothed confidence gradient is negative and the magnitude exceeds a preset threshold.

[0015] Preferably, the vehicle controller module is specifically configured to calculate the requested safe envelope radius based on the multidimensional state coupling model as follows: calculate the basic braking distance based on the current driving speed; calculate the navigation risk factor based on the smooth navigation confidence and smooth confidence gradient; calculate the network latency factor based on the network latency; calculate the load inertia factor based on the load quality data; and product and couple the navigation risk factor, network latency factor, and load inertia factor with the basic braking distance to generate the requested safe envelope radius.

[0016] Preferably, the calculation logic of the multidimensional state coupling model includes: the navigation risk factor is inversely proportional to the smooth navigation confidence, and a gradient penalty term is added when the smooth confidence gradient indicates a deterioration in positioning quality; the network latency factor is configured to map the time dimension of network latency to a buffer distance in the spatial dimension; the load inertia factor is configured to compensate the braking distance according to the incremental ratio of load quality data to empty vehicle quality; through the coupling of the above factors, the requested safety envelope radius increases with the increase of network latency, the increase of load quality data, or the deterioration of positioning quality.

[0017] Preferably, before executing the asymmetric negotiation strategy, the robot scheduling module is specifically configured to: calculate the straight-line distance between the geometric centers of the two conflicting autonomous mobile robots; determine that there is spatial overlap when the straight-line distance is less than the sum of the requested safety envelope radii of the two autonomous mobile robots; calculate the comprehensive capability score of each autonomous mobile robot based on the smooth navigation confidence, network latency, and load quality data in the vehicle state vector; the comprehensive capability score is positively correlated with the smooth navigation confidence and negatively correlated with the network latency and load quality data.

[0018] Preferably, the asymmetric negotiation strategy executed by the robot scheduling module specifically includes: comparing the comprehensive capability scores of the conflicting parties, determining the party with the higher comprehensive capability score as the first target robot, and determining the party with the lower comprehensive capability score as the second target robot; issuing an envelope maintenance command to the second target robot, allowing the second target robot to maintain its original requested safe envelope radius; issuing an envelope compression command to the first target robot, forcibly setting the safe envelope radius of the first target robot to a limit safe radius; the limit safe radius is equal to the straight-line distance minus the requested safe envelope radius of the second target robot and a preset rigid isolation distance.

[0019] Preferably, the vehicle controller module is further configured with speed space interlock logic, specifically configured as follows: when the envelope compression command is received, the upper limit of speed under the confined space is calculated in reverse based on the limit safety radius and the current load mass data; the upper limit of speed is determined based on the functional relationship between the maximum mechanical braking deceleration under the current load and the limit safety radius; and the drive module is controlled to forcibly limit the driving speed to within the upper limit of speed.

[0020] Preferably, the vehicle controller module is further configured to handle smooth transition control at the boundary of the navigation area: when a navigation source signal switching request is detected, the pose of the old navigation source at the switching moment is locked, and a virtual guide point is generated along the current velocity vector direction in combination with the data of the inertial measurement unit; a transition weight function is constructed based on the confidence change of the new navigation source; the transition weight function is used to perform weighted smooth fusion of the virtual guide point and the pose calculated by the new navigation source until the confidence of the new navigation source reaches a stable threshold.

[0021] Preferably, the transition weight function adopts an S-shaped growth curve; in the initial stage of switching, the system mainly relies on the virtual guide point for inertial extrapolation control; as the confidence of the new navigation source increases, the system gradually increases the weight of the pose calculated by the new navigation source until it is completely switched to the new navigation source.

[0022] This invention provides an intelligent scheduling system for AMR (Autonomous Mobile Robotics) workshop logistics based on composite navigation. It has the following beneficial effects:

[0023] 1. This invention utilizes the kinematic data from the inertial measurement unit through the vehicle-mounted controller module to interpolate and time-compensate the heterogeneous sampling data from the LiDAR and vision camera, and combines this with a network clock server to unify the time base of the entire system. This mechanism solves the pose calculation deviation caused by inconsistent sensor sampling frequencies, eliminates control oscillations during high-speed motion, and ensures the alignment of the feedback signals input to the control algorithm in the spatiotemporal dimension, thereby improving the positioning accuracy and motion stability of the autonomous mobile robot in dynamic environments.

[0024] 2. This invention constructs a dynamic envelope generation model based on communication-physics-sensing coupling, uniformly mapping navigation confidence gradient, network communication latency, and physical load quality to spatial geometric dimensions. The system can automatically expand the requested safety envelope radius when the sensing environment deteriorates, network latency increases, or load inertia increases. This avoids the limitations of traditional methods that only divide the safety zone based on speed, ensuring that the system reserves sufficient physical buffer space under complex operating conditions, preventing safety accidents caused by performance fluctuations of a single subsystem.

[0025] 3. This invention employs a cloud-based asymmetric negotiation and conflict decision-making strategy. Based on a comprehensive score of a vehicle's perception confidence, network response speed, and load inertia calculation capabilities, it distinguishes between the dominant and subordinate vehicles. The scheduling system maintains the original safety envelope of the subordinate vehicle unchanged, forcibly allocating spatial constraint pressure to the dominant vehicle with stronger control capabilities, ensuring its operation within the extreme safety radius. This strategy avoids the efficiency losses caused by all vehicles needing to significantly decelerate or stop in multi-vehicle encounter scenarios, maximizing traffic efficiency in confined spaces while ensuring the safety of vulnerable nodes. Attached Figure Description

[0026] Figure 1 This is a schematic diagram of the system architecture of the present invention;

[0027] Figure 2 This is a schematic diagram of the method flow of the present invention;

[0028] Figure 3 This is a schematic diagram of the multidimensional state-coupled dynamic envelope calculation model of the present invention.

[0029] Among them, 110 is the manufacturing execution module; 120 is the robot scheduling module; 130 is the communication network module; 131 is the wireless access point; 132 is the network clock server; 200 is the autonomous mobile robot; 210 is the vehicle controller module; 220 is the lidar; 230 is the bottom vision camera; 240 is the inertial measurement unit; 250 is the load detection sensor; 260 is the wireless communication module; and 270 is the drive module. Detailed Implementation

[0030] The technical solutions in 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.

[0031] See attached document Figure 1 , Figure 1 This is a schematic diagram of a system architecture according to an embodiment of the present invention. The present invention provides an AMR workshop logistics intelligent scheduling system based on composite navigation. The system mainly includes an information management layer, a control scheduling layer, and a field execution layer; the information management layer is responsible for parsing and issuing production tasks; the control scheduling layer is responsible for global path planning and traffic control; and the field execution layer is responsible for specific material handling operations.

[0032] The information management layer is mainly composed of the manufacturing execution module 110. The manufacturing execution module 110 interfaces with the upper-level enterprise resource planning system to obtain production plan information. At the same time, the manufacturing execution module 110 establishes a communication connection with the lower-level control and scheduling layer to decompose the production plan into specific logistics and handling task instructions. These task instructions include the starting position, target position, and task priority information of the materials.

[0033] The control and scheduling layer includes a robot scheduling module 120 and a communication network module 130. The robot scheduling module 120 is deployed on the server side and is configured with a global dynamic map database. The robot scheduling module 120 receives task instructions from the manufacturing execution module 110 and performs task allocation, path planning, and multi-robot collaborative traffic control based on the real-time status of each mobile robot.

[0034] The communication network module 130 adopts a wireless local area network or 5G network architecture, covering the entire workshop logistics operation area; the communication network module 130 includes several wireless access points 131 distributed in the workshop and a network clock server 132; the network clock server 132 is used to provide a unified clock reference for the entire system to ensure time synchronization of data interaction; the robot scheduling module 120 interacts with the field execution layer at high frequency through the communication network module 130.

[0035] The on-site execution layer includes several autonomous mobile robots 200; the autonomous mobile robot 200 is the physical carrier for performing logistics handling tasks and can perform autonomous positioning and navigation in the workshop environment; the autonomous mobile robot 200 is equipped with an on-board controller module 210; the on-board controller module 210 is the core computing unit of the robot, responsible for sensor data acquisition, edge-side state calculation and motion control command generation.

[0036] The autonomous mobile robot 200 is equipped with a composite navigation module for sensing environmental features and calculating vehicle pose. The composite navigation module includes a lidar 220 and a bottom vision camera 230. The lidar 220 is installed at the front or top of the robot body to scan the environmental contours and reflective pillar features, providing positioning data based on laser scanning matching. The bottom vision camera 230 is installed at the bottom of the robot body to collect QR code image data laid on the ground, providing positioning data based on vision decoding.

[0037] The vehicle controller module 210 is also connected to an inertial measurement unit 240 and a load detection sensor 250. The inertial measurement unit 240 integrates an accelerometer and a gyroscope to provide acceleration and angular velocity information of the vehicle and assist in kinematic state estimation. The load detection sensor 250 is installed at the lifting mechanism or drive module of the robot to acquire the robot's load mass data in real time.

[0038] The autonomous mobile robot 200 establishes a connection with the communication network module 130 through the wireless communication module 260; the vehicle controller module 210 monitors the quality of the communication link with the wireless access point 131 in real time through the wireless communication module 260 and obtains network latency parameters; the vehicle controller module 210 is also connected to the robot's wheel system through the drive module 270 and controls the robot's movement according to the planned path and speed commands.

[0039] In the system operation architecture, the vehicle controller module 210 performs edge computing tasks; the vehicle controller module 210 uses the data from the inertial measurement unit 240 to perform time alignment and fusion of heterogeneous data from the lidar 220 and the bottom vision camera 230 to calculate the robot's real-time pose and navigation confidence; at the same time, the vehicle controller module 210 combines the load data collected by the load detection sensor 250 and the network latency data fed back by the wireless communication module 260 to generate a vehicle state vector containing multi-dimensional features.

[0040] The robot scheduling module 120 acts as a cloud-based decision center, receiving state vectors and safety envelope requests reported by each autonomous mobile robot 200. Based on global map information, the robot scheduling module 120 performs conflict detection on multi-vehicle paths and generates targeted scheduling instructions based on an asymmetric negotiation strategy. The scheduling instructions are sent to the corresponding autonomous mobile robots 200 through the communication network module 130, and are parsed by the on-board controller module 210 to drive the drive module 270 to execute the corresponding actions, thereby realizing the collaborative scheduling of workshop logistics.

[0041] See attached document Figure 2 , Figure 2This is a schematic flowchart of an intelligent scheduling method for AMR workshop logistics based on composite navigation, according to an embodiment of the present invention. The present invention provides an intelligent scheduling method for AMR workshop logistics based on composite navigation, the method comprising the following steps:

[0042] The S100 and vehicle-mounted controller module collect heterogeneous sensor data from LiDAR and bottom vision camera. Based on a unified time reference and vehicle kinematics model for the entire system, the heterogeneous sensor data is time-aligned and weighted to calculate the real-time pose of the autonomous mobile robot.

[0043] S200 and the vehicle controller module calculate the navigation confidence at the current moment based on the matching residual of the sensor data, and smooth the first derivative of the navigation confidence to generate a smooth confidence gradient that reflects the trend of positioning stability.

[0044] The S300 and vehicle controller module calculate the requested safety envelope radius of the autonomous mobile robot based on a multi-dimensional state coupling model, according to the robot's current driving speed, smooth confidence gradient, real-time network latency fed back by the communication network module, and load quality data collected by the load detection sensor.

[0045] The S400 and vehicle controller modules generate data packets containing real-time pose, requested safety envelope radius, and vehicle state vector, and upload the data packets to the robot scheduling module through the communication network module.

[0046] S500 and the robot scheduling module receive data packets from their respective master mobile robots, detect whether there is spatial overlap in the request safety envelopes of their respective master mobile robots in the global map, and when spatial overlap is detected, execute an asymmetric negotiation strategy based on the vehicle state vector to determine the scheduling priority and target safety envelope size of the conflicting parties.

[0047] In the S600, the robot scheduling module issues differentiated envelope control commands to the conflicting parties. The on-board controller module responds to the command by adjusting the safety envelope size and simultaneously executes the speed-space interlock logic, driving the autonomous mobile robot to perform corresponding avoidance or passage actions.

[0048] The technical implementation details of the above system units and steps will be described in detail below with reference to the accompanying drawings and specific embodiments.

[0049] In this embodiment, the vehicle controller module 210 serves as the edge computing core, working in conjunction with the LiDAR 220, bottom vision camera 230, inertial measurement unit 240, and network clock server 132 to execute the following time alignment and data fusion logic:

[0050] S110, the vehicle controller module 210 establishes a time synchronization mechanism with the network clock server 132 through the wireless communication module 260; the vehicle controller module 210 maintains a system clock internally, which is periodically calibrated with the network clock server 132 to eliminate clock drift between the respective host mobile robots 200; the vehicle controller module 210 stamps the point cloud data collected by the lidar 220, the image data collected by the bottom vision camera 230, and the inertial data collected by the inertial measurement unit 240 with a global timestamp based on the unified system clock; the vehicle controller module 210 opens a sliding data window in memory and caches the most recent frames of sensor raw data in the order of timestamps.

[0051] S120, Vehicle controller module 210 determines the master clock reference time. Master clock reference time The sampling frequency depends on the control cycle of the vehicle controller module 210 or the sampling time of a sensor with a higher sampling frequency. In this embodiment, the sampling frequency of the bottom vision camera 230 is set to 30Hz, and the sampling frequency of the lidar 220 is set to 20Hz. The vehicle controller module 210 uses the trigger time of the current control cycle or the acquisition time of the latest frame of the vision image as the master clock reference time. For heterogeneous sensor data with low sampling frequency or out of sync with the master clock reference, the vehicle controller module 210 retrieves the data from the sliding data window from the time relative to the master clock reference. The most recent historical sampled data frame is denoted as the acquisition time of that historical sampled data frame. The corresponding original pose data is .

[0052] S130, the vehicle-mounted controller module 210 uses the high-frequency kinematic data collected by the inertial measurement unit 240 to perform motion compensation interpolation on the heterogeneous sensor data; the inertial measurement unit 240 outputs the instantaneous linear acceleration and angular velocity of the autonomous mobile robot 200 in the vehicle coordinate system; the vehicle-mounted controller module 210 obtains the instantaneous linear acceleration and angular velocity of the autonomous mobile robot 200 in the time period through integral calculation. Average linear velocity vector within and mean angular velocity The onboard controller module 210, based on the vehicle's kinematics model, calculates the time difference of the autonomous mobile robot 200. The pose increment generated within.

[0053] Corrected alignment pose The calculation formula is as follows:

[0054] ;

[0055] Among them, pose increment The calculation method under the two-dimensional planar motion model is as follows:

[0056] ;

[0057] In the formula: This indicates that the autonomous mobile robot 200 is in the time period The average linear velocity modulus within; This indicates that the autonomous mobile robot 200 is at time 200. The heading angle; This indicates that the autonomous mobile robot 200 is in the time period The average angular velocity within; and These represent the displacement compensation amounts in the X and Y directions in the global coordinate system, respectively. This indicates the angular compensation amount for the heading angle.

[0058] S140, the vehicle controller module 210 performs weighted fusion using the aligned pose data; for the LiDAR 220 and the bottom vision camera 230, after processing in step S130, the data of both have been uniformly converted to the master clock reference time. The equivalent state is as follows: the vehicle controller module 210 assigns weight coefficients based on the navigation area attributes where the autonomous mobile robot 200 is located; when in an area rich in reflective pillar features, the weight assigned to the LiDAR data is greater; when in an area with densely distributed QR codes, the weight assigned to the visual camera data is greater; the final fused pose It is a linear weighted sum of the aligned poses of each sensor.

[0059] Through the above steps, the vehicle controller module 210 utilizes the high-frequency characteristics of the inertial measurement unit 240 to fill the data gaps in the time axis of the low-frequency sensors, ensuring that the pose feedback signals input to the control algorithm are strictly aligned in the time dimension during the high-speed movement of the autonomous mobile robot 200, thereby eliminating the oscillation of control commands caused by asynchronous sensor sampling. As for the specific hardware selection of the inertial measurement unit 240 and the integral drift elimination algorithm, those skilled in the art can use conventional Kalman filtering or complementary filtering techniques to implement them, which will not be elaborated here.

[0060] In this embodiment, the vehicle controller module 210 not only determines the reliability of positioning based on the quality of sensor data at the current moment, but also identifies the risk of navigation degradation in advance by analyzing the time-domain change trend of confidence. The vehicle controller module 210 performs the following steps:

[0061] S210, the vehicle controller module 210 calculates the instantaneous navigation confidence based on the matching residual of the sensor measurement data; the vehicle controller module 210 acquires the laser scanning matching residual output by the laser navigation algorithm. and the QR code reprojection error output by the visual navigation algorithm Among them, laser scanning matching residuals Characterized by the average Euclidean distance deviation between the current frame's laser point cloud data and the pre-set environment map contour; QR code reprojection error It represents the deviation between the pixel coordinates of the QR code corner points observed by the camera and the theoretical pixel coordinates calculated based on the map pose.

[0062] The vehicle controller module 210 uses a negative exponential decay model to normalize the aforementioned physical errors into dimensionless confidence probability values; instantaneous navigation confidence. The calculation formula is as follows:

[0063] ;

[0064] In the formula: and These are the basic weight coefficients for the LiDAR and the bottom vision camera, respectively, and they satisfy... ; This is the standard deviation normalization parameter for the laser matching error, used to adjust the sensitivity of the laser confidence level to the error. This is the standard deviation normalization parameter for visual reprojection error, used to adjust the sensitivity of visual confidence to error; The value range is [0,1]. The closer the value is to 1, the more reliable the positioning result at the current moment.

[0065] S220, the vehicle controller module 210 performs low-pass filtering on the instantaneous navigation confidence score to generate a smooth navigation confidence score; however, the presence of reflective strips, oil stains, or temporary dynamic obstructions on the workshop floor can affect the instantaneous navigation confidence score. High-frequency random jumps occur; in order to prevent the control logic from being falsely triggered due to noise interference, the vehicle controller module 210 introduces a first-order recursive low-pass filter to smooth the original confidence sequence.

[0066] Smooth navigation confidence The calculation formula is as follows:

[0067] ;

[0068] In the formula: The smoothing confidence level of the previous control cycle; These are the filter coefficients, and their values ​​are usually set to (0, 1). The smaller the value, the stronger the suppression effect on high-frequency noise and the smoother the output curve.

[0069] S230, the vehicle controller module 210 calculates the time gradient of the smooth navigation confidence; the vehicle controller module 210 performs a differential operation on the smooth navigation confidence sequence in the time domain to obtain the rate of change of the positioning quality; the rate of change can reflect whether the autonomous mobile robot 200 is heading towards a sparse navigation feature area or leaving a reliable positioning area.

[0070] Smoothing confidence gradient The calculation formula is as follows:

[0071] ;

[0072] In the formula: The time window length for gradient calculation, which is greater than or equal to the control period; for Historical smoothed confidence levels prior to the specified time.

[0073] S240, the vehicle controller module 210 predicts navigation status trends based on smooth confidence gradients; the vehicle controller module 210 not only monitors the current absolute confidence value, but also monitors the gradient in real time. Direction and amplitude; the vehicle controller module 210 has a preset gradient degradation threshold. ( ).

[0074] When the condition is met At this time, the onboard controller module 210 determines that the autonomous mobile robot 200 is in a "navigation warning state"; at this time, even if the current moment is... If the positioning quality is still higher than the minimum positioning threshold allowed by the system, the vehicle controller module 210 will also recognize the trend of rapid deterioration of positioning quality and trigger the subsequent safety envelope adjustment logic in advance. This gradient-based trend prediction mechanism can reserve the reaction time required for deceleration or path adjustment in advance before the autonomous mobile robot 200 actually enters the signal blind zone or causes positioning divergence, thereby avoiding the system response lag problem caused by relying solely on static threshold judgment.

[0075] Reference Appendix Figure 3 In this embodiment, the vehicle controller module 210 no longer treats the autonomous mobile robot 200 as a rigid body of fixed size, nor does it generate the obstacle avoidance area based solely on the single dimension of driving speed. Instead, the vehicle controller module 210 combines the link status feedback from the wireless communication module 260 and the physical load status feedback from the load detection sensor 250 to construct a multi-dimensional state coupling model, and calculates in real time the requested safety envelope radius that can cover system hysteresis risk and physical braking risk. .

[0076] S310, the vehicle controller module 210 acquires multi-dimensional input parameters for envelope calculation; these input parameters include: the current driving speed obtained based on odometer readings or a fusion algorithm. The smooth navigation confidence score output by the aforementioned navigation confidence assessment module. and its gradient The round-trip latency of communication is monitored in real time by the wireless communication module 260. ; and the current on-board load mass collected by the load detection sensor 250. Among them, communication round-trip delay Characterized by the time difference between the data packet being sent from the onboard controller module 210 and the receipt of an acknowledgment response from the robot scheduling module 120; onboard load mass This represents the sum of the mass of the autonomous mobile robot 200 and the mass of the cargo it is currently carrying.

[0077] S320, the on-board controller module 210 calculates the basic braking distance based on the current driving speed; basic braking distance The minimum physical space required to ensure safe braking under ideal communication conditions, no load, and completely accurate positioning is represented by the following formula:

[0078] ;

[0079] In the formula: This is the braking coefficient, which is related to the response time of the servo drive system and the maximum deceleration of the motor. The current linear velocity module of the autonomous mobile robot 200; This is a static safety margin, used to cover the geometric radius of the vehicle body itself and the minimum safety clearance.

[0080] S330, the on-board controller module 210 calculates navigation risk factors based on smooth navigation confidence and its gradient; navigation risk factors This is used to correct the basic braking distance to address spatial uncertainties caused by positioning errors. When positioning accuracy decreases or shows a deteriorating trend, the system actively expands the safety envelope to offset the collision risks that may be caused by position drift. The formula for calculating the navigation risk factor is as follows:

[0081] ;

[0082] In the formula: This is the normalized smooth navigation confidence score, with a value range of (0,1]. To smooth the confidence gradient; This is a step function, taking a value of 1 when the input variable is greater than or equal to 0, and 0 otherwise; here it is set to only take a value when the gradient is negative (i.e., ...). The gradient-based penalty term is only triggered when the positioning quality deteriorates. This is the confidence level inverse weighting coefficient, used to adjust the impact of absolute positioning error on the envelope; These are gradient weight coefficients used to adjust the impact of the localization degradation trend on the envelope.

[0083] S340, the on-board controller module 210 calculates the network delay factor based on the communication round-trip time delay; network delay factor This is used to compensate for distance loss caused by control loop lag. Since the emergency stop or avoidance commands issued by the robot scheduling module 120 need to be transmitted via wireless network, a high network latency means that by the time the command reaches the on-board controller module 210, the vehicle has already passed the original safety point. Therefore, the time-dimensional delay needs to be mapped to a spatial-dimensional buffer distance. The formula for calculating the network latency factor is as follows:

[0084] ;

[0085] In the formula: For real-time monitoring of communication round-trip latency; This is the network latency sensitivity coefficient, whose dimension is the reciprocal of velocity. It is used to convert time units into dimensionless multipliers or directly relate to distance compensation.

[0086] S350, the on-board controller module 210 calculates the load inertia factor based on the vehicle load mass; load inertia factor This is used to compensate for the increased braking distance caused by the increase in physical mass; under the same braking torque, the fully loaded autonomous mobile robot 200 has greater kinetic energy and requires a longer deceleration stroke. The formula for calculating the load inertia factor is as follows:

[0087] ;

[0088] In the formula: The current real-time onboard load quality; The empty vehicle weight of the autonomous mobile robot 200; The rated maximum load capacity of the autonomous mobile robot 200; This is the load inertia weighting coefficient, used to characterize the degree of impact of load rate on braking performance degradation.

[0089] S360, the vehicle controller module 210 generates the final requested safety envelope radius based on a multi-dimensional state coupling model; the vehicle controller module 210 multiplies and couples the calculated navigation risk factor, network latency factor, and load inertia factor with the basic braking distance to obtain virtual bubble parameters including the geometric center and radius; requested safety envelope radius. The coupling calculation formula is as follows:

[0090] ;

[0091] Through the above steps, the system will determine the network status of the communication layer ( ), mechanical properties of the physical layer ( ) and the algorithm quality of the perception layer ( The unified mapping is to the geometric dimensions of the spatial layer; when the network is slow, the vehicle is heavily loaded, or the positioning is inconsistent, It will be significantly larger than the theoretical value calculated solely based on speed; this coupling mechanism ensures that even the robot scheduling module 120, based on this... Sufficient physical buffer space has been reserved during route planning or traffic control to absorb performance fluctuations of each subsystem, thereby avoiding safety accidents caused by failure of a single factor under complex operating conditions; the vehicle controller module 210 will calculate the... The data is encapsulated into a state vector and periodically reported to the robot scheduling module 120 via the wireless communication module 260, serving as the input basis for subsequent cloud-based collaborative scheduling.

[0092] In this embodiment, the robot scheduling module 120, as the decision center for multi-machine collaboration, no longer adopts a single traffic control logic. Instead, it dynamically evaluates the control capabilities of each vehicle in the confined space based on the multi-dimensional status data reported by each autonomous mobile robot 200, and performs asymmetric right-of-way allocation and envelope control accordingly.

[0093] S510, the robot scheduling module 120 performs multi-vehicle state vector analysis and spatial conflict detection; the robot scheduling module 120 periodically receives state data packets reported by all autonomous mobile robots 200 in the workshop through the communication network module 130; the data packet contains the vehicle's unique identifier ID and the real-time pose after time alignment. The request security envelope radius calculated based on the multidimensional coupling model. And a state vector containing navigation confidence, network latency, and load quality. .

[0094] The robot scheduling module 120 updates the position and virtual bubble range of each vehicle in real time on the global dynamic map; the system traverses vehicle nodes currently in the same alleyway or intersection area and calculates the Euclidean distance between each pair; when two autonomous mobile robots (denoted as...) are detected... and A path conflict is determined to exist between two objects if the following spatial overlap conditions are met:

[0095] ;

[0096] In the formula, This represents the real-time straight-line distance between the geometric centers of the two vehicles. and These are the current requested safety envelope radii for the two vehicles, respectively.

[0097] S520, the robot scheduling module 120 calculates the comprehensive capability score of conflicting vehicles based on state vectors; to resolve conflicts, the system does not adopt random waiting or first-come-first-served strategies, but rather quantitatively evaluates each vehicle's ability to "compress its own safety space and maintain stable control" at the current moment; comprehensive capability scoring function. The design logic is to award higher scores to vehicles with high perception accuracy, fast network response, and low physical inertia. (Comprehensive capability score) The calculation formula is as follows:

[0098] ;

[0099] In the formula: An index representing the vehicles involved in the conflict; For vehicles The confidence level of smooth navigation; the larger the value, the more reliable the perception. For vehicles The communication round-trip delay is used as the denominator; the lower the delay, the higher the score. For vehicles The total load mass is used as the denominator; the smaller the mass (the smaller the inertia), the higher the score. , , These are the normalization coefficients for the weights of each dimension, configured according to the emphasis on communication, sensing, or physical characteristics in the actual scenario. , is the dimensional equilibrium constant.

[0100] S530, the robot scheduling module 120 determines the superior and inferior parties based on the comprehensive capability score; the system compares the scores of the conflicting parties. and The vehicle with the higher score is designated as the "advantageous vehicle," indicating that it has stronger environmental perception and more agile control response capabilities, and can withstand more stringent motion constraints. The vehicle with the lower score is designated as the "disadvantageous vehicle," indicating that it is currently in a state of weak perception, high latency, or high inertia, and requires more lenient physical tolerance.

[0101] Assumption Then determine As the dominant party, The party at a disadvantage.

[0102] S540, the robot scheduling module 120 generates and issues asymmetric collaborative scheduling instructions; the system adopts a differentiated space allocation strategy based on comprehensive capability scoring to maximize the space utilization of the channel.

[0103] For the disadvantaged side The robot scheduling module 120 issues an "envelope preservation" command; the system allows... Continue to maintain the larger security envelope of its original request. This means that the disadvantaged party does not need to bear the risk of collision caused by the compressed space in order to avoid it, and still has enough buffer distance to offset network lag or braking inertia.

[0104] For the advantageous party The robot scheduling module 120 issues an "envelope compression" command; the system forcibly calculates a new, smaller limit safety radius. To adapt to the current physical spacing. Advantageous side limit safety radius. The calculation formula is as follows:

[0105] ;

[0106] In the formula, The minimum physical rigidity separation distance set for the system is used to prevent the two vehicles from coming into contact.

[0107] Through this calculation, the robot scheduling module 120 prioritizes allocating the available free space between the two vehicles to the disadvantaged vehicle in a worse state, while applying spatial constraint pressure to the advantageous vehicle in a better state; the robot scheduling module 120 will include Control commands are sent to the dominant party via communication network module 130. The vehicle-mounted controller module 210; Advantages Upon receiving the instruction, the local electronic fence parameters will be updated immediately, and this limit radius will be used as the new constraint boundary in subsequent motion control. This asymmetric negotiation mechanism ensures that in mixed fleets, complex network environments, and differentiated load operation scenarios, the logistics system can fully explore the control potential of high-performance nodes (advantageous parties) while ensuring the safety of the weakest node (disadvantageous party). This avoids the efficiency loss caused by the "weakest link effect" in traditional scheduling, which requires all vehicles to slow down or stop and wait.

[0108] In this embodiment, the vehicle controller module 210 and the drive module 270 work together to convert the scheduling instructions sent from the cloud into underlying physical motion constraints and handle the smooth transition control at the junction of different navigation areas.

[0109] S610, the on-board controller module 210 executes the speed-space interlock logic; when the autonomous mobile robot 200, as the dominant party, receives the envelope compression command issued by the robot scheduling module 120, its allowed geometric safe area is forcibly reduced to the limit safe radius. To ensure physical braking within the confined space, the vehicle controller module 210 must reverse-calculate the maximum permissible driving speed based on the current physical load status; the vehicle controller module 210 reads data from the load detection sensor 250 in real time to obtain the current total load mass. It also queries the pre-stored braking performance curve to determine the maximum available deceleration under the current load. .

[0110] The vehicle controller module 210 calculates the upper limit of speed in a confined space based on kinematic formulas. The calculation formula is as follows:

[0111] ;

[0112] In the formula: The compressed limit safety radius issued by robot scheduling module 120; This refers to the vehicle's static safety margin, which includes the vehicle's physical radius and minimum rigid stopping clearance. This represents the maximum mechanical braking deceleration under the current load mass. The function is used to ensure that the value inside the square root is non-negative.

[0113] The vehicle controller module 210 will calculate the The highest priority hard constraint command is issued to the drive module 270; when executing the motion control command generated by the path planning algorithm, the drive module 270 compares the target linear velocity with the target linear velocity in real time. If the target linear velocity exceeds The drive module 270 forcibly clamps the output speed to... Within this range; the interlock mechanism ensures that the geometry of the safety envelope is always dynamically matched with the vehicle's physical kinetic energy, preventing insufficient braking distance due to insufficient speed reduction caused by space compression.

[0114] S620, the vehicle controller module 210 performs inertial extrapolation smooth transition control at the boundary of the navigation area; when the autonomous mobile robot 200 moves from the laser navigation area into the QR code navigation area (or vice versa), the vehicle controller module 210 will detect a switching request for the navigation source signal; in the initial stage of switching, due to the small number of observations or poor geometric distribution of the new navigation source (such as a QR code that has just entered the field of view), its corresponding navigation confidence is low. Fluctuations are possible.

[0115] To avoid control divergence, the vehicle controller module 210 initiates inertial extrapolation logic; the vehicle controller module 210 locks the high-confidence pose output by the old navigation source (e.g., LiDAR) from the previous moment. Combined with the short-time integral data from the inertial measurement unit 240, the vehicle's current velocity vector direction is measured at a preset forward distance. Generate a virtual boot point Virtual guide point The calculation formula is as follows:

[0116] ;

[0117] In the formula: This is the trigger time for the switching action; This is the instantaneous velocity vector calculated for inertial navigation. The time constant for virtual guidance is used to determine the look-ahead of the virtual point.

[0118] S630, the vehicle controller module 210 performs weighted pose smoothing fusion; during inertial extrapolation, the vehicle controller module 210 monitors the confidence level of the new navigation source in real time. The onboard controller module 210 constructs a transition weight function that varies with time or confidence level. Control feedback pose during transition period The calculation is as follows:

[0119] ;

[0120] Transition weight function The calculation uses an S-shaped growth curve, and the specific formula is as follows:

[0121] ;

[0122] In the formula: The confidence level of the new navigation source at the current moment; The threshold is half the confidence level (inflection point). The slope factor is used to control the steepness of the weight changes.

[0123] When the confidence level of the new navigation source At lower levels, Approaching 0, the system primarily relies on virtual boot points for control; as... Rise and exceed the threshold The value quickly approaches 1, achieving a complete switch to the new navigation source; if it occurs within the preset timeout period... If the stable threshold is not reached, the vehicle controller module 210 triggers degraded braking, the control drive module 270 stops the motor output, and reports a positioning anomaly alarm to the robot scheduling module 120. Through this inertial extrapolation mechanism, the system uses the short-term high-precision characteristics of inertial navigation as a pose reference benchmark to guide the vehicle to smoothly pass through the boundary area where the signal is unstable, eliminating the path change or positioning jump that may be caused by traditional hard handover.

[0124] Although embodiments of the invention have been shown and described, it will be understood by those skilled in the art that various changes, modifications, substitutions and alterations can be made to these embodiments without departing from the principles and spirit of the invention, the scope of which is defined by the appended claims and their equivalents.

Claims

1. An AMR workshop logistics intelligent scheduling system based on composite navigation, characterized in that, It includes an information management layer, a control and scheduling layer, and an on-site execution layer; The information management layer includes a manufacturing execution module; the control and scheduling layer includes a robot scheduling module and a communication network module, the communication network module including a network clock server that provides a unified clock reference for the entire system; the field execution layer includes an autonomous mobile robot, the autonomous mobile robot being equipped with a composite navigation module, a wireless communication module, a drive module, and an on-board controller module; the composite navigation module includes a lidar, a bottom vision camera, an inertial measurement unit, and a load detection sensor. The vehicle controller module is configured to use the time reference of the network clock server to perform time alignment and weighted fusion of heterogeneous sensor data from the LiDAR and bottom vision camera using the kinematic data of the inertial measurement unit, and calculate the real-time pose and navigation confidence. The vehicle controller module is also configured to calculate the requested safety envelope radius based on a multi-dimensional state coupling model according to the robot's current driving speed, the trend of navigation confidence, the network latency monitored by the wireless communication module, and the load quality data collected by the load detection sensor, and upload the data packet containing the requested safety envelope radius and the vehicle state vector to the robot scheduling module. The robot scheduling module is configured to detect whether there is spatial overlap between the requested safety envelopes of each host mobile robot, and when spatial overlap is detected, calculate the comprehensive capability score of each conflicting party based on the vehicle state vector, execute an asymmetric negotiation strategy to determine the target safety envelope size of each conflicting party, and issue envelope control commands. Specifically, the asymmetric negotiation strategy executed by the robot scheduling module includes: comparing the comprehensive capability scores of the conflicting parties, identifying the party with the higher comprehensive capability score as the first target robot, and the party with the lower comprehensive capability score as the second target robot; issuing an envelope maintenance command to the second target robot, allowing the second target robot to maintain its original requested safety envelope radius; and issuing an envelope compression command to the first target robot, forcibly setting the safety envelope radius of the first target robot to a limit safety radius. The limit safety radius is equal to the straight-line distance between the geometric centers of the conflicting parties minus the requested safety envelope radius of the second target robot and a preset rigid isolation distance. The onboard controller module is also configured to adjust the safety envelope size in response to an envelope control command and drive the autonomous mobile robot to perform actions; wherein, the onboard controller module is also configured with speed space interlock logic, specifically configured to: when receiving the envelope compression command, reversely calculate the upper speed limit value under the confined space based on the limit safety radius and the current load mass data; the upper speed limit value is determined based on the functional relationship between the maximum mechanical braking deceleration under the current load and the limit safety radius; the control drive module forcibly limits the driving speed to within the upper speed limit value.

2. The AMR workshop logistics intelligent scheduling system based on composite navigation according to claim 1, characterized in that, The vehicle controller module is specifically configured to perform time alignment and weighted fusion of heterogeneous sensor data as follows: The system clock is synchronized with the network clock server via the wireless communication module, and a sliding data window is created in memory to cache the raw sensor data with a global timestamp. Determine the master clock reference time and retrieve historical sampling data frames from the asynchronous sensor from the sliding data window; The linear acceleration and angular velocity output by the inertial measurement unit are used for integration to calculate the pose increment from the historical sampling time to the master clock reference time; The pose increment is compensated to the original pose data corresponding to the historical sampling data frame to obtain the aligned pose, and the aligned pose of each sensor is weighted and fused according to the current navigation area attributes.

3. The AMR workshop logistics intelligent scheduling system based on composite navigation according to claim 1, characterized in that, The vehicle controller module is specifically configured to calculate the trend of navigation confidence changes as follows: Instantaneous navigation confidence is calculated based on the matching residuals of sensor measurement data, and smoothed using a low-pass filter to generate smooth navigation confidence. The smoothed navigation confidence is calculated by performing a difference operation in the time domain to determine the smoothed confidence gradient; The direction and magnitude of the smooth confidence gradient are monitored in real time. When the smooth confidence gradient is negative and the magnitude exceeds a preset threshold, it is determined that the positioning quality is deteriorating.

4. The AMR workshop logistics intelligent scheduling system based on composite navigation according to claim 3, characterized in that, The vehicle controller module is specifically configured to calculate the requested safety envelope radius based on the multidimensional state coupling model as follows: Calculate the basic braking distance based on the current driving speed; The navigation risk factor is calculated based on the smooth navigation confidence level and the smooth confidence level gradient. Calculate the network latency factor based on the aforementioned network latency; Calculate the load inertia factor based on the load quality data; The navigation risk factor, network latency factor, and load inertia factor are productized and coupled with the basic braking distance to generate the requested safety envelope radius.

5. The AMR workshop logistics intelligent scheduling system based on composite navigation according to claim 4, characterized in that, The computational logic of the multidimensional state coupling model includes: The navigation risk factor is inversely proportional to the smooth navigation confidence, and a gradient penalty term is added when the smooth confidence gradient indicates a deterioration in positioning quality; The network latency factor is configured to map the time dimension of network latency to a buffer distance in the spatial dimension; The load inertia factor is configured to compensate for braking distance based on the incremental ratio of load mass data to empty vehicle mass. The coupling of the above factors causes the requested security envelope radius to increase with increasing network latency, increased load quality data, or deterioration of location quality.

6. The AMR workshop logistics intelligent scheduling system based on composite navigation according to claim 1, characterized in that, The robot scheduling module is specifically configured as follows before executing the asymmetric negotiation strategy: Calculate the straight-line distance between the geometric centers of the two conflicting autonomous mobile robots; When the straight-line distance is less than the sum of the requested safety envelope radii of the two autonomous mobile robots, it is determined that there is spatial overlap; The comprehensive capability score of each autonomous mobile robot is calculated based on the smooth navigation confidence, network latency, and load quality data in the vehicle state vector. The overall capability score is positively correlated with smooth navigation confidence and negatively correlated with network latency and load quality data.

7. The AMR workshop logistics intelligent scheduling system based on composite navigation according to claim 1, characterized in that, The on-board controller module is also configured to handle smooth transition control at the boundary of navigation areas: When a navigation source signal switching request is detected, the pose of the old navigation source at the switching moment is locked, and a virtual guide point is generated along the current velocity vector direction in combination with the data from the inertial measurement unit. A transition weight function is constructed based on the change in confidence of the new navigation source; The poses calculated by the virtual guide point and the new navigation source are weighted and smoothly fused using the transition weight function until the confidence level of the new navigation source reaches a stable threshold.

8. The AMR workshop logistics intelligent scheduling system based on composite navigation according to claim 7, characterized in that, The transition weight function adopts an S-shaped growth curve; in the initial stage of switching, the system mainly relies on the virtual guide point for inertial extrapolation control; as the confidence of the new navigation source increases, the system gradually increases the weight of the pose calculated by the new navigation source until it is completely switched to the new navigation source.

Images