Omnidirectional intelligent mobile forklift control system

Abstract

The invention relates to the technical field of warehouse logistics equipment, and discloses an omni-directional intelligent mobile forklift control system. The system comprises a dynamic path planning module, an omnidirectional motion control module, a load stability monitoring module, an environment adaptive adjustment module and an intelligent obstacle avoidance decision module. The dynamic path planning module collects storage environment multi-dimensional spatial data and constructs a dynamic environment topological map; the omni-directional motion control module analyzes driving wheel parameters based on the map and generates an omni-directional motion composite motion instruction set; the load stability monitoring module collects a fork pressure distribution matrix and a cargo gravity center offset according to an instruction set, and calculates a stability compensation parameter; the environment adaptive adjustment module fuses illumination and slope data to generate a motion control constraint condition; and the intelligent obstacle avoidance decision module predicts a dynamic obstacle trajectory, outputs an obstacle avoidance path re-planning instruction and updates a map. The system can meet the requirements of modern intelligent warehousing.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of warehouse logistics equipment, in particular to an omnidirectional intelligent mobile forklift control system. BACKGROUND

[0002] In the modern warehouse logistics system, the efficiency of cargo handling and the safety of operation directly affect the operation efficiency of the overall supply chain, and the performance level of the forklift as the core handling equipment plays a key role in the quality of warehouse operation. The current mainstream forklift control system adopts a fixed path planning mode. When building an environmental map, such a system usually only collects two-dimensional coordinates of the shelves and static obstacle position information, and is difficult to cope with dynamic changes in the warehouse environment, such as temporary personnel, handling equipment, or fine-tuning of shelf positions. When these dynamic factors appear, the forklift with fixed path planning often needs manual intervention to adjust the path, which not only affects the operation efficiency, but also may cause safety hazards due to human judgment errors. The motion control of the existing forklifts mostly adopts a single driving mode, which is difficult to achieve omnidirectional movement. In narrow warehouse passageways, turning and direction adjustment operations are complex and require a large operating space. This has become an important factor restricting the efficiency of warehouse operations in the context of increasing warehouse space utilization. In addition, in the process of cargo handling, the existing system has a simple monitoring of load stability, usually only by detecting the lifting height of the forks to make a judgment, ignoring situations such as offset of the center of gravity of the cargo and uneven distribution of the load surface pressure. When the forklift encounters uneven ground, slope changes, or acceleration and deceleration during movement, the center of gravity of the cargo is likely to shift, which may cause the cargo to tilt and damage, or even fall, threatening the safety of personnel and equipment. The current forklift control system has weak self-adaptability to the environment and cannot adjust the motion control strategy according to environmental parameters such as light intensity, ground friction coefficient, and slope in different warehouse areas. For example, in a warehouse area with dim light, the sensors of the forklift may have reduced data collection accuracy due to insufficient light; in an area with low ground friction coefficient, the forklift may skid if it still travels at the regular speed; and in an area with a slope, the lack of targeted torque distribution may result in insufficient power or excessive energy consumption. The existence of these problems makes the existing forklift control system difficult to meet the needs of modern intelligent warehouses in terms of operation efficiency, safety, and adaptability, and there is an urgent need for an intelligent mobile forklift control system that can cope with dynamic environments, achieve omnidirectional movement, ensure load stability, and have environmental self-adaptability. SUMMARY

[0003] The present application aims to provide an omnidirectional intelligent mobile forklift control system to solve the problems raised in the background.

[0004] In order to achieve the above object, the application provides a kind of omnidirectional intelligent mobile forklift control system, the system includes: The dynamic path planning module obtains multi-dimensional space data of the warehouse environment, constructs a dynamic environment topology map by real-time collection of shelf three-dimensional coordinates, obstacle contour point cloud and ground friction coefficient; The omnidirectional motion control module is based on the dynamic environment topology map, analyzes the independent steering angle of each drive wheel of the forklift chassis and the torque distribution ratio, generates a composite motion instruction set for omnidirectional movement; The load stability monitoring module collects the pressure distribution matrix of the fork bearing surface and the gravity center offset according to the composite motion instruction set, and calculates the stability compensation parameter in the dynamic handling process; The environment adaptive adjustment module generates motion control constraint conditions for different warehouse areas based on the stability compensation parameter, and fuses current environment light intensity and ground slope data; The intelligent obstacle avoidance decision module performs probabilistic prediction on the motion trajectory of dynamic obstacles according to the motion control constraint conditions, outputs obstacle avoidance path re-planning instructions and updates the dynamic environment topology map.

[0005] Preferably, the dynamic path planning module comprises: The space data acquisition submodule synchronously acquires the geometric feature point cloud of the shelf column through laser radar and depth camera, and extracts the calibration parameters of the shelf height and channel width; The topology modeling submodule converts the point cloud data into a rasterized node network according to the calibration parameters, and labels the topological connection relationship of the passable area and the fixed obstacle; The path optimization submodule calculates the shortest path node sequence from the starting point to the target point based on the rasterized node network using a bidirectional search algorithm, generates an initial navigation path and writes it into the dynamic environment topology map.

[0006] Preferably, the omnidirectional motion control module comprises: The kinematic analysis submodule decomposes the coupling relationship between the four-wheel independent steering angle of the forklift chassis and the drive wheel speed according to the curvature change point of the initial navigation path; The torque distribution submodule calculates the torque distribution weight of each drive wheel in the acceleration stage based on the coupling relationship, and generates the composite motion instruction set containing the steering angle increment and the torque adjustment threshold.

[0007] Preferably, the load stability monitoring module comprises: The pressure sensing submodule acquires the real-time pressure value of each monitoring point on the fork bearing surface through an array type pressure sensor, and constructs a two-dimensional pressure distribution matrix; The gravity center calculation submodule calculates the lateral and longitudinal offset of the gravity center of the cargo relative to the central axis of the forks based on the two-dimensional pressure distribution matrix and the three-dimensional scanning data of the cargo shape contour; The compensation generation submodule derives the overturning moment compensation coefficient of the forklift when turning based on the lateral and longitudinal offset, and outputs the stability compensation parameter.

[0008] Preferably, the environment adaptive adjustment module comprises: The light analysis submodule collects intensity distribution data of the warehouse area ceiling light source, and identifies the navigation positioning error correction amount of the low-illumination area; The slope compensation submodule obtains the pitch angle and roll angle of the ground slope through the inertial measurement unit, and calculates the influence factor of gravity on cargo slip; The constraint generation submodule fuses the navigation positioning error correction amount and the influence factor to generate the motion control constraint condition of limiting the maximum speed of omnidirectional movement and the steering angle rate.

[0009] Preferably, the intelligent obstacle avoidance decision module comprises: The trajectory prediction submodule identifies the motion direction and velocity vector of the dynamic obstacle through a multi-target tracking algorithm, and predicts the collision probability distribution in a future time window; The path update submodule labels high-risk areas in the dynamic environment topological map according to the collision probability distribution and re-plans the obstacle avoidance path node sequence, and outputs the obstacle avoidance path re-planning instruction.

[0010] Preferably, the system further comprises: The shelf identification calibration module scans the texture features of the shelf signboard through the laser radar, and matches the target shelf position data in the pre-stored shelf number database; The dynamic path planning module corrects the end point coordinates of the initial navigation path according to the target shelf position data.

[0011] Preferably, the system further comprises: The battery management module monitors the remaining capacity and output current fluctuation of the forklift power battery in real time, and predicts the sustainable running time in the current task cycle; The dynamic path planning module filters the reachable shelf operation area in the dynamic environment topological map according to the sustainable running time.

[0012] Preferably, the system further comprises: The fault diagnosis module collects temperature and vibration spectrum data of each drive motor, and identifies the fault feature mode of abnormal wear parts; The omnidirectional motion control module dynamically disables the abnormal drive wheel and adjusts the torque distribution strategy of the composite motion instruction set according to the fault feature mode.

[0013] Preferably, the system further comprises: The wireless communication module receives a multi-fork truck cooperative task instruction set issued by a central dispatching system. The intelligent obstacle avoidance decision module labels the real-time positions and planned paths of other fork trucks in the dynamic environment topology map according to the multi-fork truck cooperative task instruction set.

[0014] Compared with the prior art, the present application has the following advantages: The omnidirectional intelligent mobile forklift control system acquires multi-dimensional spatial data of the warehouse environment through a dynamic path planning module, collects real-time three-dimensional coordinates of shelves, obstacle contour point clouds, and ground friction coefficients, and constructs a dynamic environment topology map, which can accurately capture dynamic changes in the warehouse environment. Whether it is a temporary obstacle or a slight adjustment of the shelf position, it can be perceived in time and integrated into map updating, without the need for manual intervention to achieve dynamic adaptation of the path, so that the forklift always maintains a reasonable driving path in a complex and variable warehouse environment, reduces work interruptions caused by environmental changes, and improves overall warehouse handling efficiency. The omnidirectional motion control module analyzes the independent steering angles and torque distribution ratios of each drive wheel of the forklift chassis based on the dynamic environment topology map, generates a composite motion instruction set for omnidirectional movement, breaks through the limitations of the traditional single drive mode of the forklift, and enables the forklift to achieve flexible omnidirectional movement. Whether it is lateral movement, in-place rotation, or diagonal driving in a narrow warehouse passageway, it can be easily completed, greatly reducing the forklift's demand for operating space, effectively improving the utilization rate of warehouse space, and significantly reducing the time and space required for forklift steering and turning, further optimizing the work process. The load stability monitoring module collects the pressure distribution matrix of the fork bearing surface and the center of gravity offset of the goods in real time according to the composite motion instruction set, calculates the stability compensation parameters in the dynamic handling process, and monitors the load state from multiple dimensions, not limited to traditional fork height detection. It can accurately capture potential risks such as center of gravity offset of goods and uneven pressure on the bearing surface. The stability compensation parameters calculated can adjust the motion state of the forklift in time, such as adjusting the driving speed, steering angle, or torque distribution when detecting the center of gravity offset of the goods, to avoid tilting, falling, and other situations during handling, ensuring the safety of the goods handling process, reducing economic losses caused by goods damage, and also reducing personnel and equipment safety risks caused by goods falling. The environmental self-adaptive adjustment module fuses the current environmental light intensity and ground slope data based on the stability compensation parameter to generate motion control constraint conditions for different warehouse areas, so that the forklift has strong environmental adaptability. In the area with insufficient light intensity, the accuracy of data acquisition can be ensured by adjusting the sensitivity of the sensor or the auxiliary lighting strategy; in the area with low ground friction coefficient, the sliding phenomenon can be avoided by restricting the driving speed and optimizing the torque distribution; in the area with slope, the torque distribution of the driving wheel can be adjusted according to the slope data to ensure that the forklift has sufficient power when climbing and maintains stable driving when descending, avoiding the problems of high energy consumption or insufficient power, so that the forklift can maintain a stable and efficient working state in different environmental conditions of the warehouse area. The intelligent obstacle avoidance decision module can not only identify dynamic obstacles, but also predict their motion trajectories, judge potential collision risks in advance, output obstacle avoidance path re-planning instructions in time, and ensure that the forklift can quickly and safely adjust the path when encountering dynamic obstacles to avoid collision accidents. At the same time, the obstacle avoidance path re-planning instructions can also update the dynamic environment topology map in real time, providing accurate environmental data for subsequent path planning, forming a virtuous cycle, and further improving the safety and efficiency of forklift operation. BRIEF DESCRIPTION OF DRAWINGS

[0015] Figure 1 The timing diagram of the omnidirectional intelligent mobile forklift control system described in the present application; Figure 2 The working principle diagram of the dynamic path planning module; Figure 3 The working principle diagram of the omnidirectional motion control module. DETAILED DESCRIPTION

[0016] The technical solutions in the embodiments of the present application will be described clearly and completely below with reference to the drawings in the embodiments of the present application. Obviously, the described embodiments are only part of the embodiments of the present application, not all the embodiments. Based on the embodiments in the present application, all other embodiments obtained by those skilled in the art without creative labor are within the scope of protection of the present application.

[0017] Please refer to Figure 1The application provides an omnidirectional intelligent mobile forklift control system, which comprises the cooperation of multiple modules. A dynamic path planning module collects multidimensional spatial data such as the three-dimensional coordinates of shelves, obstacle contour point clouds and ground friction coefficients in a warehouse environment in real time through a laser radar, a depth camera and a friction coefficient sensor. After filtering and fusion processing, the data are used to construct a dynamic environment topology map. The map represents the connection relationship between passable areas and fixed obstacles in a rasterized form. An omnidirectional motion control module receives the dynamic environment topology map as input, analyzes the independent steering angles and torque distribution ratios of each drive wheel of the forklift chassis, and calculates a composite motion instruction set containing a steering sequence and a speed curve through a kinematic model. A load stability monitoring module triggers an array-type pressure sensor on the fork bearing surface according to the composite motion instruction set, collects a pressure distribution matrix and calculates the center of gravity offset of the goods in combination with three-dimensional scanning data, so as to deduce stability compensation parameters in the dynamic handling process. An environment adaptive adjustment module integrates an illumination sensor and an inertial measurement unit to obtain environmental illumination intensity and ground slope data, and generates motion control constraint conditions for low-illumination or slope areas after fusing the stability compensation parameters. An intelligent obstacle avoidance decision module predicts the motion trajectory probability of a dynamic obstacle based on the motion control constraint conditions by using a multi-target tracking algorithm, outputs obstacle avoidance path re-planning instructions and updates the dynamic environment topology map in real time, and forms a closed-loop control.

[0018] Example 1: see Figure 2, the laser radar and the depth camera are integrated on the protective bracket of the forklift, and data acquisition processes are started simultaneously in a hardware-synchronized manner. The laser radar adopts a rotating mechanical structure, and the laser beams emitted thereby are continuously scanned in a horizontal plane for 360 degrees, and a certain elevation angle range is covered in the vertical direction through a multi-beam layout, so as to capture three-dimensional point cloud data of the shelf column. The depth camera cooperates with the light-emitting diode to project a speckle pattern, and the depth value of each pixel point is calculated by using the binocular vision principle. The data of the two sensors are sent to a preprocessing unit after being aligned by time stamp. The preprocessing of the point cloud data includes denoising, registration and segmentation steps. The denoising algorithm uses a statistical outlier removal method to filter out interference points caused by dust or flying insects. In the registration process, the laser radar point cloud and the depth camera point cloud are aligned by an iterative closest point algorithm to form a dense point cloud model in a unified coordinate system. The segmentation algorithm extracts the continuous surface of the shelf column based on a region growing technique, and fits the geometric features such as the edge line and the corner point of the column. The calibration parameters of the shelf height and the passage width are automatically calculated from the segmented point cloud. The height is obtained by detecting the point cloud clustering height difference of the adjacent shelf beams, and the passage width is determined by analyzing the minimum passable distance between the point clouds of the two side shelf columns. These parameters are recorded as metadata and stored together with the point cloud. The topological modeling submodule receives the preprocessed point cloud data, first projects the three-dimensional point cloud onto a two-dimensional horizontal plane to generate an occupancy grid map, and sets the size of each grid to 0.1 meters square according to the size of the forklift body and the safety margin. The grid value is marked as free, occupied or unknown according to the point cloud density.

[0019] The grid-based map is converted into a node network graph, each free grid center point is taken as a node, and the connection relationship between the nodes is established based on the eight-neighborhood or four-neighborhood rule to form the vertices and edges of the graph structure. The grids corresponding to fixed obstacles such as shelf bases or walls are marked as impassable nodes, and their connection edges are removed. The calculation of the topological connection relationship adopts a graph traversal algorithm such as breadth-first search to verify the reachability between nodes and specially mark dead ends or narrow passages, and generates a map representation in the form of an adjacency matrix or adjacency list. The network graph is dynamically updated to reflect environmental changes. The path optimization submodule initializes the node numbers corresponding to the start point and the target point based on the constructed grid-based node network, and simultaneously expands the search from the start point node and the target point node by using a bidirectional search algorithm. The open list and the closed list are maintained for the two search directions respectively, and the open list stores the nodes to be expanded, and the closed list records the visited nodes. Each node evaluates its heuristic cost to the start point and the target during the search process, and the cost function considers the Euclidean distance and the turning penalty factor. When the open lists of the two search directions have an intersection node, the algorithm terminates and traces back the path. The generated node sequence is converted into a continuous coordinate point sequence to form an initial navigation path, which contains position points and recommended directions, and is written into the path layer of the dynamic environmental topology map for calling by other modules.

[0020] The laser radar scanning frequency of the spatial data acquisition submodule is set to 10 Hz, the point cloud density is configurable to tens of thousands of points per frame, the resolution of the depth camera is maintained at 1920x1080 pixels, the exposure time is adapted to the ambient light intensity, and the point cloud registration uses a quaternion method to calculate the rotation and translation matrix. When extracting the calibration parameters, the calculation of the shelf height is determined by histogram analysis of the vertical direction of the point cloud to determine the peak interval, the channel width uses a convex hull algorithm to detect the boundary point cloud of the obstacles on both sides, and then calculates the nearest pair distance. All parameters are smoothed by Kalman filtering to reduce jitter. The gridding processing of the topological modeling submodule uses a sliding window method to update the local map in real time, a unique identifier is assigned to each node when constructing the node network, and the weight of the connection edge is adjusted according to the actual distance between the grids and the difficulty of passing, such as increasing the weight in the slope area. The verification of the topological connection relationship calculates the shortest path of all node pairs by the Floyd-Warshall algorithm to ensure global reachability analysis, and the labeling of fixed obstacles combined with historical data avoids misjudgment. When implementing the bidirectional search algorithm of the path optimization submodule, the heuristic function uses Manhattan distance to accelerate convergence, the turning penalty factor is set according to the minimum turning radius of the forklift, and the node sequence is smoothed by B-spline interpolation during path backtracking to eliminate jagged paths.

[0021] The coordinate point sequence of the initial navigation path is attached with timestamp and confidence information, and the update mechanism of the dynamic environment topology map uses a differential update strategy to modify only the changed area to reduce the computational load. The hardware synchronization of the spatial data acquisition submodule is achieved by a GPS clock signal with microsecond-level precision, the point cloud denoising uses radius filtering to remove isolated points, the registration algorithm uses feature point matching to improve efficiency, and the geometric feature extraction uses principal component analysis to calculate the normal vector and curvature of the point cloud. The calculation results of the calibration parameters are stored in a ring buffer to support multi-frame data fusion to improve robustness. The grid size of the topological modeling submodule can be dynamically adjusted according to the complexity of the environment, virtual nodes are added to the node network graph to handle dynamic obstacles, and the topological connection relationship uses a union-find data structure to maintain connected components to ensure the integrity of the path search. The search algorithm of the path optimization submodule automatically inserts intermediate nodes in areas with sparse path points to avoid control difficulties of long straight-line paths, and performs a collision detection loop after generating the path to ensure safety and feasibility.

[0022] Example 2: see Figure 3, the kinematics analysis submodule receives the initial navigation path from the path planning module, which is composed of a series of continuous path points with curvature information. The analysis process first discretizes the path, extracting key path points at intervals of the forklift control period. Each key point contains global coordinates, orientation angle, and path curvature data. The forklift chassis uses a four-wheel independent steering and driving Mecanum wheel structure. The kinematics parameters of each wheel are calculated by establishing the transformation relationship between the vehicle body coordinate system and the global coordinate system. The analysis algorithm is based on rigid body kinematics, which decomposes the overall motion velocity vector of the vehicle body into the linear velocity and steering angle requirement of each wheel. The processing of curvature change points is particularly critical, as the differential speed and steering angle change sequence of each wheel need to be calculated in advance on high-curvature road segments. The coupling relationship analysis considers the motion constraints between the wheel groups and describes the mapping relationship between the vehicle body speed and the wheel speed through the Jacobian matrix. The matrix elements include the wheel installation position geometric parameters and the instantaneous steering angle value. The analysis result generates the theoretical steering angle and speed reference value of each drive wheel at each point on the path. The torque distribution submodule calculates the dynamic weight based on the coupling relationship output by the kinematics analysis. The weight distribution is based on the adhesion coefficient estimate and the actual load of each wheel. The load information is determined by the vehicle body inclination sensor and air spring pressure data.

[0023] The torque distribution in the acceleration phase uses a feedforward and feedback control structure. The feedforward part calculates the basic torque demand based on the acceleration command and the wheel moment of inertia, and the feedback part corrects the torque output through motor speed closed loop. The distribution algorithm ensures that the slip rate of each wheel is maintained within the optimal range. The generation process of the composite motion command set integrates the steering angle increment and the torque adjustment threshold. The steering angle increment is represented in differential form as the angle change between adjacent control periods. The torque adjustment threshold is set based on the torque upper limit dynamically calculated by the motor thermal model and the battery discharge capacity. The command set is sent to each in-wheel motor controller through the CAN bus at a fixed period. The pressure sensing submodule of the load stability monitoring module manages 32 pressure sensing units on the fork bearing surface through an embedded system. The sensing units are arranged in a matrix to cover the main bearing area of the fork. Each unit contains a temperature-compensated strain gauge and a signal conditioning circuit. The data acquisition period is synchronized with the motion control period. The two-dimensional pressure distribution matrix is constructed using an interpolation algorithm to enhance spatial resolution. The matrix element values are normalized to represent the relative pressure intensity of each region. The matrix data is updated in real time and marked with a time stamp.

[0024] The center of gravity calculation submodule synchronously acquires three-dimensional scanning data of the cargo shape contour, which is collected by a three-dimensional TOF camera installed on the portal frame. The point cloud data is processed by plane fitting and contour extraction algorithm to obtain the circumscribed cube model of the cargo. The center of gravity offset calculation combines the pressure distribution matrix and the cargo geometric model, and adopts a multi-sensor fusion algorithm to weight the pressure distribution center of gravity and the geometric center. Finally, the lateral and longitudinal offset of the cargo center of gravity relative to the fork center axis is output, and the offset is expressed in the form of vector to represent the size and direction. After the center of gravity offset of the compensation generation submodule is input, the dynamics model of the forklift-cargo system is first established, which includes the vehicle mass parameters, cargo mass parameters and suspension system characteristics. The overturning moment calculation considers the coupling effect of centrifugal force generated by steering acceleration and gravity of the cargo. The compensation coefficient derivation is based on the moment balance equation, and the theoretical stable boundary is compared with the actual measured value to generate compensation parameters, including hydraulic system pressure adjustment value, speed limit suggestion value and steering angular velocity correction value. The output format of the stability compensation parameter is a structured data packet, which includes compensation action time window, parameter priority flag and failure protection condition. The path curvature calculation of the kinematics analysis submodule adopts the differential geometry method, and the curvature radius threshold is set in association with the minimum turning radius of the forklift. The analytical algorithm automatically inserts a transition curve for sharp turning paths to ensure motion smoothness. The weight calculation of the torque distribution submodule introduces an adaptive algorithm, which adjusts the distribution ratio in real time according to the motor current feedback. The torque adjustment threshold is set with double threshold values to prevent overload. The signal conditioning circuit of the pressure sensing submodule includes an anti-aliasing filter and a programmable gain amplifier. After digital filtering of the sampling value, mechanical vibration interference is eliminated. The fusion algorithm of the center of gravity calculation submodule adopts Kalman filter to reduce measurement noise, and the spatio-temporal registration of the geometric model and pressure data is realized through a hardware synchronization signal.

[0025] The dynamic model parameters of the compensation generation submodule can be calibrated on site by a calibration program, and the compensation parameters are subjected to a rate of change limiting process before output to avoid sharp jumps. The data exchange between the omnidirectional motion control module and the load stability monitoring module is realized through shared memory, the control period is unified to 10 milliseconds, and a heartbeat detection mechanism is established between modules to ensure communication reliability. The verification of the composite motion instruction set is carried out in a hardware-in-the-loop test in a simulation environment, and an instruction cache mechanism is used to ensure control continuity in the presence of network delays. The calibration of the pressure sensor matrix uses standard weight blocks for linearity calibration, and the temperature drift compensation coefficients are stored in non-volatile memory. The calculation of the center of gravity offset introduces a confidence evaluation, and when the confidence is low, it automatically switches to a conservative control strategy. The stability compensation parameters are integrated with the vehicle stability control system to form a multi-layer safety protection architecture. The kinematics analysis submodule supports smooth transition processing when the path is re-planned, and automatically calculates the interpolated trajectory after the path point is updated to avoid motion mutations. The slip rate control of the torque distribution submodule uses a model predictive control algorithm to predict the change in wheel-ground adhesion in advance. The fault detection function of the pressure sensor matrix can identify the failure of a single sensor and reconstruct the pressure distribution through adjacent sensor data. The three-dimensional point cloud processing of the center of gravity calculation submodule optimizes the point cloud density, reducing the computational load while ensuring accuracy. The roll moment calculation of the compensation generation submodule introduces a road slope compensation factor to improve the stability control accuracy in slope conditions.

[0026] In embodiment 3, the light analysis submodule of the environment self-adaptive adjustment module continuously collects ambient light data through an array of light intensity sensors installed on the top of the forklift, which consists of eight light-sensitive elements arranged in a ring layout. Each element has a sampling frequency set to 20 Hz. After analog-to-digital conversion, the data is transmitted to a signal processor. The processor applies a Gaussian filter algorithm to smooth the original illumination values to eliminate transient fluctuations. The identification of low-light areas is based on a pre-set threshold comparison. When the average illumination is below 50 lux, it is marked as a low-light area. The calculation of the navigation positioning error correction quantity combines the measurement deviations of the vision sensor and the laser radar. The Kalman filter is used to estimate the positioning drift caused by changes in illumination. The correction quantity is used to adjust the reference input of the path tracking controller in real time in differential form. The slope compensation submodule relies on a high-precision inertial measurement unit to monitor ground inclination changes. The unit contains a three-axis accelerometer and a gyroscope, which output pitch and roll angle data at a frequency of 200 Hz. These data are combined with wheel speed meter information through a complementary filtering algorithm to calculate the tilt angle of the vehicle body relative to the horizontal plane. The influence factor of gravity on cargo slip is derived using a physical model, taking into account the static friction coefficient between the cargo and the surface of the forks and the load weight. The influence factor calculation combines real-time kinetic parameters to dynamically assess the stability risk of the cargo on the slope. The constraint generation submodule integrates the navigation positioning error correction quantity output by the light analysis submodule and the influence factor provided by the slope compensation submodule. A fuzzy logic control system is used for multi-source information fusion. The input variables of the fuzzy system include illumination error, slope angle, load weight, etc. The output variables are the limiting values of the maximum speed and steering angular rate of omnidirectional movement. These constraints are adaptively adjusted according to environmental changes to form a set of motion control constraints.

[0027] The trajectory prediction submodule of the intelligent obstacle avoidance decision module processes dynamic obstacle data from sensors through a multi-target tracking algorithm. The tracking algorithm is based on a detection-association framework. First, the YOLOv4 model is used to identify moving objects such as people or other forklifts in the image. Then, the Kalman filter is used to predict the motion direction and velocity vector of each obstacle. The prediction time window is set to 5 seconds, and the future trajectory point sequence is generated at intervals of 0.1 seconds. The calculation of collision probability distribution uses a logistic regression model, where the collision probability is defined by the following formula:

[0028] where: P represents the collision probability, with a value range of 0 to 1, is a sensitivity coefficient that controls the steepness of the probability curve, is the real-time distance between the forklift and the obstacle, is a distance threshold, which is dynamically adjusted according to the forklift speed and the type of obstacle. The path updating submodule monitors the collision probability distribution in real time. When the probability exceeds 0.7, it is marked as a high-risk area and marked with a red grid in the dynamic environment topology map. The path re-planning uses the RRT* algorithm, which randomly expands the tree structure from the current path point until a new path node sequence is found that bypasses the high-risk area. The new path is smoothed by B-spline curve to ensure executability, and finally outputs the obstacle avoidance path re-planning instruction to the control system.

[0029] The light intensity sensor array of the light analysis submodule is calibrated periodically by a standard light source. The response curve of each element is stored in non-volatile memory. The illumination threshold is dynamically adjusted according to the warehouse operating time, for example, the threshold is reduced to 30 lux at night to adapt to weak environmental light conditions. The application of navigation positioning error correction uses a gradual adjustment strategy to avoid sudden changes that can cause unstable control. The correction data is strictly synchronized with the timestamp of the positioning system. The inertial measurement unit data fusion algorithm of the slope compensation submodule introduces wheel speedometer mileage information to correct integral drift. The pitch and roll angle calculation results are low-pass filtered to remove high-frequency noise. The calculation of the gravity influence factor refers to the forklift tilt sensor data in real time to ensure that the model matches the actual working conditions. The fuzzy logic control system rule base of the constraint generation submodule contains dozens of empirical rules, such as "if the illumination is low and the slope is large, then significantly reduce the speed". The defuzzification method uses the barycentric method to output accurate constraint values. The multi-target tracking algorithm of the trajectory prediction submodule supports obstacle ID maintenance and handles occlusion through appearance feature matching. The state vector of the Kalman filter includes position, velocity, and acceleration. The confidence of the predicted trajectory is evaluated based on historical tracking accuracy. The RRT* algorithm of the path updating submodule adaptively adjusts the number of sampling points according to the complexity of the environment, increases the sampling density in narrow passages, and considers the minimum turning radius constraint of the forklift in path smoothing to ensure physical feasibility. Data exchange between modules is realized through a high-speed bus with a period of 10 milliseconds to ensure real-time performance.

[0030] In the depth implementation of the light analysis submodule, the layout of the light intensity sensor array optimizes the coverage of the 180-degree view in the forward direction of the forklift, the spectral response of each sensor element matches the characteristics of artificial light sources, and the data acquisition circuit includes a programmable gain amplifier to adapt to different light ranges; the window size of the Gaussian filter is dynamically adjusted according to the speed of the forklift, and the window size is reduced at high speed to improve the response speed. The navigation positioning error correction model introduces an ambient light color temperature compensation factor, as different light source color temperatures affect the exposure parameters of the vision sensor, and the correction amount calculation combines inertial navigation system data to improve robustness. The installation position of the inertial measurement unit of the slope compensation submodule is dynamically balanced and calibrated to reduce the influence of vehicle body vibration on angle measurement, and the calculation of pitch angle and roll angle uses quaternion representation to avoid gimbal lock problems; the calculation of the gravity influence factor integrates the hydraulic pressure data of the forks to estimate the friction state of the goods and forks in real time, and the influence factor output is a dimensionless value with a range of 0 to 1, representing the slip risk level.

[0031] The input variables of the fuzzy logic system of the constraint generation submodule are normalized, the illumination error is mapped to the [0, 1] interval, the slope angle is converted to an angle value, and the de-fuzzified output constraint value is subjected to amplitude limiting to prevent extreme cases. The set of motion control constraints is sent to the chassis controller through the CAN bus, and the instruction format includes a timestamp and a priority flag. The YOLOv4 model of the trajectory prediction submodule is trained using a warehouse scene dataset to optimize the detection accuracy of forklifts and personnel, and the process noise and observation noise parameters of Kalman filtering are adaptively adjusted online. Parallel computing is used to accelerate the calculation of collision probability distribution to meet real-time requirements. The RRT* algorithm of the path update submodule is implemented with heuristic guidance to improve search efficiency, and collision detection is performed in the path verification stage until safety is achieved. The obstacle avoidance path replanning instruction includes path point sequences and control parameters to ensure smooth transitions. The cooperation between the environment adaptive adjustment module and the intelligent obstacle avoidance decision module is coordinated through a central scheduler, and the module state monitoring heartbeat mechanism is used to switch to a safe mode in case of an exception. The entire system implementation focuses on real-time performance and reliability to meet the complex operation requirements of intelligent forklifts. The data preprocessing of the light analysis submodule adds an outlier detection algorithm to exclude transient light source interference; the influence factor calculation of the slope compensation submodule considers dynamic load changes, such as weight distribution adjustment when goods are shaken; the fuzzy rule of the constraint generation submodule supports online learning to optimize rule weights based on historical operation data. The parameters of the collision probability formula of the trajectory prediction submodule are calibrated through actual measurement data, and the sensitivity coefficient λ is set to different values according to the type of obstacles, with λ = 0.5 for personnel and λ = 0.3 for forklifts; the replanning result of the path update submodule is verified for stability to ensure that the new path meets the dynamics constraints. The communication protocol between modules uses a custom binary format to reduce transmission delay, and data integrity is guaranteed through CRC verification.

[0032] Embodiment 4: Collaborative workflow of shelf recognition calibration module and battery management module, its operation process is revealed through a specific task scenario: assume that the forklift receives the instruction issued by the central dispatching system, requiring to go to the shelf numbered A-05-12 for picking work. The shelf recognition calibration module first starts the laser radar to scan the surrounding environment, the laser beam is emitted at a frequency of 10 Hz and the return signal is received, after the point cloud data is denoised by voxel filtering, the possible shelf column structure is segmented by region growing algorithm. When the metal structure with signboard is detected, the scanning focus is automatically adjusted to the signboard area, the point cloud density is increased to twice the standard value to capture more rich surface details. The texture feature extraction of the signboard uses the local binary pattern algorithm, which converts the point cloud intensity value into a rotation-invariant texture feature code, and the feature code is matched with the pre-stored shelf number database. The database is stored in the vehicle-mounted solid state disk, recording the global coordinates, height specifications and special operation requirements corresponding to each shelf number. The matching process uses the approximate nearest neighbor search algorithm, when the similarity of the feature code and the database record A-05-12 exceeds the threshold value, the system determines that the target shelf is successfully identified. At this time, the module reads the accurate coordinates of the shelf stored in the database (X=125.63m, Y=88.47m, Z=5.21m), and takes this coordinate as the new endpoint of path planning. After receiving the target shelf position data, the dynamic path planning module immediately starts the path correction program.

[0033] The correction algorithm first calculates the deviation vector of the current navigation path endpoint and the original target point, and then iteratively adjusts the path node sequence using the gradient descent method. During the iteration process, the adjustment amount of each path point is inversely proportional to its distance from the new endpoint, the closer the point, the larger the adjustment amplitude, ensuring smooth transition of the path. The corrected path endpoint coordinate accuracy is controlled within ±2 cm, and the path curvature is recalculated to ensure compliance with the forklift kinematic constraints. The whole correction process is completed within 200 milliseconds, realizing seamless path switching. The monitoring function of the battery management module is realized by high-precision coulomb meter, which samples the total current and single voltage of the battery in real time, with a sampling frequency of 1 kHz. After the current data is filtered by sliding window average, the remaining power is estimated combined with the battery temperature compensation model. The output current fluctuation analysis uses fast Fourier transform to identify the main harmonic components and their energy distribution. The sustainable running time prediction is based on the dynamic power consumption model, which considers the current load weight of the forklift, the slope change of the planned path and the historical energy consumption data. The prediction algorithm divides the remaining power by the average power consumption in the last five minutes to get the basic running time, and then modifies it according to the path characteristics: for example, increase the time consumption coefficient on the ascending slope section, and reduce the coefficient on the descending slope section. Referring to Table 1, the battery key parameter monitoring records during task execution are shown.

[0034] Table 1: Battery parameter monitoring and running time prediction Time stamp Remaining power (%) Average current (A) Temperature (°C) Load weight (kg) Predicted remaining time (min) 14:30:15 78.3 25.6 28.4 0 126 14:32:40 76.8 43.2 29.1 350 118 14:35:20 75.1 38.9 29.8 350 115 The dynamic path planning module performs reachable region screening in the dynamic environment topology map according to the sustainable runtime data provided by the battery management module. The screening algorithm first calculates the minimum energy consumption of the round-trip path from the current location to each rack region. The energy consumption model includes basic movement energy consumption, lifting energy consumption, and turning additional energy consumption. Then, the energy consumption value is converted into time demand and compared with the predicted remaining time. Only the rack region with a time demand less than 70% of the predicted remaining time is marked as a reachable region. This conservative threshold ensures that the forklift has sufficient power to return to the charging station. The reachable region information is overlaid on the environment map in different color layers for the path planning algorithm to preferentially select. The database update mechanism of the rack identification calibration module is designed in dual mode: in online mode, it receives incremental updates from the central database through the wireless network, and in offline mode, it relies on local cache data to maintain operation. The backup plan when the sign recognition fails includes rack outline matching and relative position calculation algorithm, which improves the system robustness. The laser radar scanning parameters are automatically adjusted according to the ambient light. In strong light conditions, the transmission power is increased to ensure the quality of the point cloud. The power consumption model parameters of the battery management module can be optimized adaptively through learning algorithms, recording the deviation between actual energy consumption and predicted value under different working conditions, and gradually correcting the model coefficients. The temperature compensation model considers the battery aging factor, and adjusts the compensation curve after more than 500 charge and discharge cycles. The sustainable runtime prediction result is updated every 30 seconds, and the system alarm is triggered when there is a significant change. The reachable region screening result of the dynamic path planning module affects the path decision in real time. When the predicted remaining time is lower than the threshold, the system automatically plans the optimal path to the charging station. The gradient descent method in the path correction process sets the maximum number of iterations to prevent infinite loop, and the convergence condition is based on the quadratic norm of coordinate error.

[0035] The signboard texture feature extraction also integrates optical character recognition technology as an auxiliary verification means. After the spatiotemporal alignment of point cloud data and visual images, the number characters on the signboard are recognized using a convolutional neural network. The database query uses a multi-level cache architecture, with the most recently accessed records stored in memory to speed up response time. The coordinate correction value is smoothed by Kalman filtering to avoid frequent path adjustments due to identification jitter. The current sampling circuit of the battery management module uses a 24-bit high-precision ADC, and the current sensor uses a closed-loop Hall effect principle to improve measurement linearity. The slope influence coefficient in the power consumption model is dynamically calculated based on inertial measurement unit data, providing real-time feedback on actual road conditions. The prediction algorithm automatically switches to a conservative mode when detecting an abnormally large current consumption, extending the predicted remaining time by 20% as a safety margin. The reachable region screening algorithm of the dynamic path planning module incorporates a dynamic obstacle influence factor, which appropriately increases the time budget when there are moving obstacles in the path. The screening results are combined with task priorities for processing, and high-priority tasks can be allowed to relax the accessibility conditions. The entire system focuses on data consistency between modules, ensuring that each module's decisions are based on the same time slice data.

[0036] In Example 5, assume that a forklift numbered AGV-03 is executing a transportation task from a picking station to a shelf A-05 in the warehouse area B2 channel. The temperature sensor of the fault diagnosis module is embedded in the winding of the four drive motors of the forklift, and a PT100 platinum resistance element is used to collect temperature data at a frequency of 10 times per second. Vibration monitoring is achieved through a three-axis accelerometer installed on the motor housing, with a sampling frequency set to 1 kHz. When AGV-03 passes through the channel curve, the temperature data of the right front wheel drive motor continuously rises from the normal working temperature of 65 degrees Celsius to 82 degrees Celsius within 30 seconds, and the vibration accelerometer detects an abnormal vibration component with a frequency of 87 Hz, which matches the bearing fault characteristic frequency. The wavelet transform analysis unit of the fault diagnosis module decomposes the vibration signal into 5 layers, and extracts a clear impact component in the third layer of detail coefficients, with a matching degree of 0.76 with the pre-stored bearing inner ring wear pattern in the fault feature library, indicating that the motor has an early fault risk. After receiving the fault feature pattern, the omni-directional motion control module immediately starts the fault-tolerant control program. The control algorithm first marks the abnormal drive wheel as disabled and fixes its steering angle through a mechanical locking device to prevent free rotation from interfering with the vehicle's attitude.

[0037] The torque distribution strategy is adjusted to three-wheel drive mode, and the torque distribution weights of the remaining three drive wheels are recalculated. The new distribution scheme takes into account the mass redistribution and steering center offset, and uses the principle of force couple balance to maintain the driving direction. The generation of the composite motion instruction set increases the restriction condition, and the maximum speed is reduced from 1.5 meters per second to 1.0 meters per second, and the steering angle rate is limited to 60% of the normal value. These adjustments are issued in real time to each motor controller through the control local area network. The wireless communication module maintains a 5G connection with the central dispatching system during this period, receiving collaborative task instruction sets containing multi-forklift position information. The instruction set is packaged in JSON format, including task priority, path reservation time window, and emergency braking protocol information. When AGV-03 enters the intersection of the shelf area, the communication module receives a warning message that AGV-07 is approaching from the side passage. This vehicle has a higher priority and carries fragile goods. The intelligent obstacle avoidance decision module immediately activates the collaborative obstacle avoidance mode and marks the real-time position (coordinates X=135.72, Y=92.45) of AGV-07 and the path point sequence of its planned path in the dynamic environment topology map.

[0038] The system uses a time window conflict detection algorithm to calculate the space-time overlap probability of the paths of the two vehicles in the intersection area and finds that there is a collision risk within the next 8 seconds. The path update sub-module starts dynamic path re-planning and generates an avoidance path for AGV-03 that includes a temporary waiting point: first, stop at the current position for 3 seconds, then follow the modified path to complete the remaining task after AGV-07 passes. The node sequence of the re-planned path is shared with AGV-07 through V2X communication to form a collaborative obstacle avoidance scheme. The deep monitoring function of the fault diagnosis module continuously enhances data acquisition after detecting an anomaly, with a temperature sampling frequency of 20Hz and envelope spectrum detection added to the vibration analysis. The severity assessment of bearing faults uses trend analysis to calculate the temperature rise rate and vibration energy growth rate, and triggers a secondary alarm when the indicators exceed the threshold. The three-wheel drive model of the fault-tolerant control system is pre-verified through offline simulation, and multiple torque distribution schemes are stored for different fault wheel positions to ensure the smoothness of the switching process. The data link of the wireless communication module establishes a double-checking mechanism, each data packet contains a sequence number and a timestamp, and automatically switches to a backup frequency band when the packet loss rate exceeds 5%. The collaborative task instruction set analysis uses a state machine model to verify the integrity of the instructions and the execution conditions, and invalid instructions are stored in the log for subsequent analysis.

[0039] The conflict detection algorithm of the intelligent obstacle avoidance decision module introduces a prediction confidence evaluation. When the positioning signal of AGV-07 fluctuates greatly, the detection time window is automatically expanded to 12 seconds to improve safety. In the specific implementation process, the bearing fault feature library of the fault diagnosis module is constructed through laboratory accelerated life testing, containing vibration spectrum templates at different wear stages. The base function of wavelet transform is Db4 wavelet, and the decomposition layer is adaptively adjusted according to the motor speed. The trend analysis of temperature data uses the exponential weighted moving average algorithm to eliminate random fluctuation interference. The fault-tolerant control switching process of the omni-directional motion control module includes a 200 millisecond transition period, during which fuzzy control is used to gradually adjust the torque distribution. The path tracking algorithm in three-wheel drive mode introduces lateral error compensation, which adjusts the rear wheel steering angle to offset the yawing moment caused by asymmetric driving force. The speed limit parameter is dynamically adjusted according to the severity of the fault, and when the temperature exceeds 85 degrees Celsius, the speed is further reduced to 0.8 meters per second. The 5G connection of the wireless communication module uses a private network frequency band, and the transmission delay is controlled within 50 milliseconds. The message queue of the cooperative task instruction set sets priority management, and the emergency avoidance instruction can interrupt the regular state reporting. The path re-planning result of the intelligent obstacle avoidance decision module is verified for stability to ensure that the forklift in the fault state still meets the dynamic constraints.

[0040] The interface between the fault diagnosis module and the vehicle management system uses a publish-subscribe mode, and fault events are broadcast to related subsystems through a message middleware. The torque distribution strategy of the omni-directional motion control module pre-stores response schemes for various fault scenarios, including single-wheel failure, double-wheel failure, and other extreme cases. The signal strength monitoring of the wireless communication module is linked with path planning, and a conservative path is planned in advance in signal blind areas. The cooperative algorithm of the intelligent obstacle avoidance decision module supports a multi-vehicle negotiation mechanism, which automatically uses a preset avoidance rule library to make decisions when detecting that the other vehicles are simultaneously avoiding. The entire implementation process embodies the balance between real-time, reliability, and safety. Early warning of fault diagnosis provides fault tolerance processing time for the system, wireless communication ensures the efficiency of multi-agent cooperation, and intelligent obstacle avoidance decision-making realizes the unity of safety and efficiency on this basis. The data flow between modules is synchronized using a unified time scale, and the execution results of control instructions are verified through sensor feedback to form a closed-loop control system.

[0041] It should be noted that, in this text, relationship terms such as first and second are only used to distinguish one entity or operation from another, and do not necessarily require or imply any actual relationship or order between the entities or operations. Moreover, the terms "include", "contain" or any other variant thereof are intended to cover non-exclusive inclusion, so that a process, method, article or device that includes a series of elements not only includes those elements, but also includes other elements not explicitly listed or inherent to such process, method, article or device.

[0042] While embodiments of the present application have been shown and described, it is to be understood that various modifications, substitutions, replacements and variations can be made to these embodiments without departing from the principles and spirit of the present application, the scope of which is defined by the appended claims and their equivalents.

Claims

1. An omnidirectional intelligent mobile forklift control system, characterized in that, Comprise: The dynamic path planning module acquires multi-dimensional space data of the warehouse environment, constructs a dynamic environment topology map by real-time collection of shelf three-dimensional coordinates, obstacle contour point cloud and ground friction coefficient; The omnidirectional motion control module analyzes the independent steering angle and torque distribution ratio of each drive wheel of the forklift chassis based on the dynamic environment topology map, and generates a composite motion instruction set for omnidirectional movement; The load stability monitoring module collects the pressure distribution matrix of the fork bearing surface and the gravity center offset according to the composite motion instruction set, and calculates the stability compensation parameter in the dynamic handling process; The environment adaptive adjustment module generates motion control constraint conditions for different warehouse areas based on the stability compensation parameter, and fuses the current environment light intensity and ground slope data; The intelligent obstacle avoidance decision module predicts the motion trajectory of dynamic obstacles according to the motion control constraint conditions, outputs obstacle avoidance path re-planning instructions and updates the dynamic environment topology map.

2. The omnidirectional intelligent mobile fork truck control system of claim 1, wherein, The dynamic path planning module comprises: The space data acquisition submodule synchronously acquires the geometric feature point cloud of the shelf column through laser radar and depth camera, and extracts the calibration parameters of the shelf height and channel width; The topology modeling submodule converts the point cloud data into a rasterized node network according to the calibration parameters, and labels the topological connection relationship of the passable area and the fixed obstacle; The path optimization submodule calculates the shortest path node sequence from the starting point to the target point based on the rasterized node network, generates an initial navigation path and writes it into the dynamic environment topology map.

3. The omnidirectional intelligent mobile fork truck control system of claim 2, wherein, The omnidirectional motion control module comprises: The kinematics analysis submodule decomposes the coupling relationship between the four-wheel independent steering angle and the drive wheel speed of the forklift chassis according to the curvature change point of the initial navigation path; The torque distribution submodule calculates the torque distribution weight of each drive wheel in the acceleration stage based on the coupling relationship, and generates the composite motion instruction set containing the steering angle increment and the torque adjustment threshold.

4. The omnidirectional intelligent mobile fork truck control system of claim 3, wherein, The load stability monitoring module comprises: The pressure sensing submodule acquires the real-time pressure value of each monitoring point on the fork bearing surface through the array type pressure sensor, and constructs a two-dimensional pressure distribution matrix; The gravity center calculation submodule calculates the lateral and longitudinal offset of the gravity center of the goods relative to the fork center axis according to the two-dimensional pressure distribution matrix and the three-dimensional scanning data of the goods contour; The compensation generation submodule derives the overturning moment compensation coefficient of the forklift when steering based on the lateral and longitudinal offset, and outputs the stability compensation parameter.

5. The omnidirectional intelligent mobile fork truck control system of claim 4, wherein, The environment adaptive adjustment module comprises: The light analysis submodule collects the intensity distribution data of the warehouse area ceiling light source, and identifies the navigation positioning error correction amount of the low illumination area; The slope compensation submodule acquires the pitch angle and roll angle of the ground slope through the inertial measurement unit, and calculates the influence factor of gravity on goods sliding; The constraint generation submodule fuses the navigation positioning error correction amount and the influence factor to generate the motion control constraint condition limiting the maximum speed of omnidirectional movement and the steering angle rate.

6. The omnidirectional intelligent mobile fork truck control system of claim 5, wherein, The intelligent obstacle avoidance decision module comprises: The trajectory prediction submodule identifies the motion direction and velocity vector of the dynamic obstacle through a multi-target tracking algorithm, and predicts the collision probability distribution in a future time window; The path update submodule labels high-risk areas in the dynamic environment topology map and re-plans the obstacle avoidance path node sequence according to the collision probability distribution, and outputs the obstacle avoidance path re-planning instruction.

7. The omnidirectional intelligent mobile fork truck control system of claim 6, wherein, Further comprising: The shelf recognition calibration module scans the texture features of the shelf identification plate through the laser radar, and matches the target shelf position data in the pre-stored shelf number database; The dynamic path planning module corrects the end point coordinates of the initial navigation path according to the target shelf position data.

8. The omnidirectional intelligent mobile fork truck control system of claim 7, wherein, Further comprising: The battery management module monitors the remaining capacity and output current fluctuation of the forklift power battery in real time, and predicts the sustainable running time in the current task cycle; The dynamic path planning module screens the reachable shelf operation area in the dynamic environment topology map according to the sustainable running time.

9. The omnidirectional intelligent mobile fork truck control system of claim 8, wherein, Further comprising: The fault diagnosis module collects the temperature and vibration spectrum data of each drive motor, and identifies the fault feature mode of the abnormal wear parts; The omni-directional motion control module dynamically disables the abnormal drive wheel and adjusts the torque distribution strategy of the composite motion instruction set according to the fault feature mode.

10. The omnidirectional intelligent mobile fork truck control system of claim 9, wherein, Further comprising: The wireless communication module receives the multi-forklift cooperative task instruction set issued by the central dispatching system; The intelligent obstacle avoidance decision module labels the real-time position and planned path of other forklifts in the dynamic environment topology map according to the multi-forklift cooperative task instruction set.

Images