Unmanned forklift indoor navigation system and navigation method

Abstract

The invention discloses an unmanned forklift indoor navigation system and method, and the method comprises the steps: obtaining navigation data of an unmanned forklift based on an indoor environment, and determining a current position and a target position of the unmanned forklift; performing dynamic path planning based on the current position and the target position to generate an optimal navigation path; according to the optimal navigation path, a control instruction is generated, and the unmanned forklift is commanded to execute corresponding navigation and operation. Through real-time path planning and dynamic adjustment, the unmanned forklift can quickly respond to environmental changes, the waiting time is shortened, and the logistics efficiency is improved.

Description

Technical Field

[0001] The present invention relates to the field of navigation technology, and in particular to an indoor navigation system and a navigation method for an unmanned forklift. Background Art

[0002] As a crucial component of intelligent logistics systems, unmanned forklifts are widely used in modern warehousing, manufacturing, and distribution centers. Their core function is to enable autonomous navigation and cargo handling, thereby reducing manual intervention and improving logistics efficiency. Typically, unmanned forklift navigation systems rely on multiple sensors and complex algorithms to achieve precise positioning, path planning, and obstacle avoidance.

[0003] Existing unmanned forklift systems primarily rely on laser navigation, vision navigation, and hybrid navigation technologies. These technologies rely on auxiliary equipment such as reflectors, QR codes, labels, or pre-placed landmarks within the work environment to enable forklift positioning and path planning in complex environments. Path planning modules in existing technologies typically rely on pre-stored maps to enable forklift navigation. When a forklift first enters a work area, manual operation is required to map the area and store the map data in the system. Subsequent navigation relies on these pre-stored static maps.

[0004] Although existing unmanned forklift navigation systems have made some progress, many shortcomings still exist in practical applications: current navigation systems are highly dependent on preset markers or static maps, and have difficulty adapting quickly to dynamically changing complex environments. For example, when the cargo stacking position, obstacle distribution, or environmental layout changes, the system cannot respond effectively and requires manual intervention to update the map. Existing sensing modules have difficulty providing high-precision environmental information in complex scenarios (such as narrow shelves or obstructed areas), resulting in low fork positioning accuracy of target cargo, which in turn affects cargo handling efficiency. The anti-collision module's perception of obstacles is mostly based on a single sensor such as lidar, which is easily affected by environmental interference (such as dust, light, or obstructions). In dynamic scenarios, the system's path adjustment is less real-time and robust, making collisions or delays more likely to occur. The path planning module usually generates a fixed path based on a static map. When faced with real-time obstacles or path congestion, the existing technology lags in path adjustment, making it difficult to achieve flexible obstacle avoidance and efficient path optimization.

[0005] Therefore, there is an urgent need for an indoor navigation system and method for an unmanned forklift. Summary of the Invention

[0006] The present invention provides an indoor navigation system and a navigation method for an unmanned forklift to solve the above-mentioned problems existing in the prior art.

[0007] In order to achieve the above object, the present invention provides the following technical solutions:

[0008] An indoor navigation method for an unmanned forklift, comprising:

[0009] S101: Obtain navigation data of the unmanned forklift based on the indoor environment to determine the current position and target position of the unmanned forklift;

[0010] S102: Perform dynamic path planning based on the current position and the target position to generate an optimal navigation path;

[0011] S103: Generate control instructions based on the optimal navigation path to instruct the unmanned forklift to perform corresponding navigation and operations.

[0012] Wherein, step S101 includes:

[0013] S1011: Acquire spatial distribution data of the indoor environment based on indoor sensors, and generate an indoor navigation topology network based on the spatial distribution data;

[0014] S1012: Based on the navigation topology network, determine the driving area and obstacle distribution of the unmanned forklift in the indoor environment and obtain real-time environmental perception data;

[0015] S1013: Based on environmental perception data, use visual SLAM technology to perform real-time mapping and positioning to determine the current precise position of the unmanned forklift and the preset target position.

[0016] Wherein, step S102 includes:

[0017] S1021: Compare the current position of the unmanned forklift with the target position, calculate the shortest path, and optimize the path curve by combining the forklift's turning radius, acceleration, and deceleration parameters;

[0018] S1022: During the path optimization process, determine whether the path passes through a narrow passage or a complex area, obtain environmental data of the area, and make detailed adjustments to the path based on the size and driving capacity of the forklift;

[0019] S1023: Based on the driving characteristics of the forklift, determine the key turning points and parking points on the path, combine the environmental data, and make detailed adjustments to these key points to generate the optimal navigation path.

[0020] Wherein, step S103 includes:

[0021] S1031: Generate motion control instructions for the unmanned forklift based on the optimal navigation path. The motion control instructions include steering, acceleration, and deceleration.

[0022] S1032: driving the unmanned forklift along the planned path by executing control instructions;

[0023] S1033: Monitor the movement status of the unmanned forklift in real time and perform path deviation correction.

[0024] Wherein, after step S103, the following steps are included:

[0025] When an unmanned forklift is transporting goods, the status of the goods is monitored in real time based on the data from the unmanned forklift's indoor sensors. This includes detecting whether the goods are damaged or unbalanced.

[0026] Analyze cargo status data and identify potential cargo transportation risks in advance;

[0027] When abnormalities are detected in the cargo, the emergency braking mechanism is triggered to ensure the safety of the cargo and the unmanned forklift.

[0028] Among them, determining the driving area and obstacle distribution of unmanned forklifts in indoor environments includes:

[0029] When the obstacle density detected in the driving area by real-time environmental perception data exceeds the preset threshold, it is determined that there is a potential obstacle risk in the current driving area, a red mark is added to the navigation topology network, and an obstacle warning notification is sent to the warehouse management system;

[0030] Otherwise, the current driving area is determined to be in normal operation, and a green mark is added to the navigation topology network;

[0031] If there are potential obstacle risks in the current driving area, including:

[0032] Obtain the real-time navigation signal of the unmanned forklift, determine the destination of the unmanned forklift based on the navigation information, and generate a recommended driving route corresponding to the destination based on the navigation topology network and real-time environmental perception data;

[0033] Obtain the expected obstacle density for each time period corresponding to each recommended route, determine whether the expected obstacle density for the time period corresponding to the predicted travel time of the undetermined driving area exceeds the preset density of the recommended driving area, and if so, regenerate the recommended driving route;

[0034] Otherwise, the recommended driving route is sent to the corresponding area of ​​the navigation display interface of the unmanned forklift for display;

[0035] After confirming that the unmanned forklift has adopted the recommended driving route, based on the current driving data of the unmanned forklift, the time period corresponding to the unmanned forklift's arrival at each section of the recommended route and its corresponding driving area is predicted to generate the recommended driving route;

[0036] Based on the recommended driving route, the expected obstacle density of each recommended driving area is marked on the navigation topology network.

[0037] Among them, judging whether the path passes through a narrow channel or a complex area includes:

[0038] Obtain path planning data and build a path optimization model;

[0039] The path optimization model is configured with a path detection window set, and each path detection window in the path detection window set is provided with a path complexity evaluation image;

[0040] Collecting path travel data, dividing the path travel data into path detection windows, and inputting the data into each path detection window for window filling;

[0041] Obtaining the window filling result and determining the path complexity based on the ratio of the filling area of ​​the path detection window to the total area of ​​the path detection window in the path optimization model;

[0042] The path optimization model includes a path detection window and a path evaluation mechanism;

[0043] The path detection window is used to import the path planning data into the corresponding path detection window according to the corresponding path segment;

[0044] The path evaluation mechanism is embedded in the path detection window and is used to compare and cover the environmental data imported into the path detection window and output the coverage area;

[0045] Route adjustment includes: obtaining route travel data;

[0046] Input the path driving data into the preset generative adversarial network for path division, and obtain the path identification output by the generative adversarial network;

[0047] The generative adversarial network consists of a generator and a dual-output discriminator;

[0048] Generative adversarial networks are trained as follows:

[0049] Obtain path planning samples, path type parameters and path import identifier loading parameters;

[0050] Inputting the path planning sample into the generator for pre-training to obtain a first recognition network output by the generator;

[0051] The path planning sample, path type parameter and path import identifier are loaded into the discriminator for training to obtain the discriminant branch of the discriminator;

[0052] Determine the path corresponding to the path planning data based on the discriminant branch and the first identification network, and generate a path identifier for importing the path data into the identification mechanism;

[0053] According to the path identification, the path planning data is imported into different path detection windows for path filling;

[0054] Also includes:

[0055] Get the filling data of each path detection window;

[0056] Predetermining the upper and lower thresholds of the path complexity corresponding to each path detection window;

[0057] Based on the upper and lower thresholds, the path data is monitored for over-limit, and the abnormal complexity is determined based on the over-limit monitoring results;

[0058] Perform anomaly evaluation on the path data according to the anomaly complexity of each path detection window to determine the path evaluation data;

[0059] Each path is automatically refined and adjusted based on the path evaluation data.

[0060] Among them, real-time monitoring of cargo status includes:

[0061] Sensors on unmanned forklifts collect real-time dynamic data on cargo, including displacement, vibration, tilt angle, and other physical characteristics related to cargo status.

[0062] Process and analyze the collected dynamic data to identify the movement state of the goods during transportation and detect whether there is abnormal vibration or tilt;

[0063] If abnormal vibration or tilt is detected, the status of the cargo is determined to be damaged or unbalanced. If so, an early warning message is generated and a processing flow is triggered;

[0064] Through the cargo status assessment model, a comprehensive assessment of the cargo status is conducted based on sensor data to determine whether the cargo is in normal transportation status. If the cargo status is abnormal, corrective measures are taken according to preset rules;

[0065] The cargo status assessment results are combined with real-time monitoring data to generate a cargo status report, and the transportation path or movement of the unmanned forklift is adjusted in real time to avoid cargo damage or imbalance problems.

[0066] Among them, an unmanned forklift indoor navigation system includes:

[0067] A navigation data acquisition unit, used to acquire navigation data of the unmanned forklift based on the indoor environment and determine the current position and target position of the unmanned forklift;

[0068] A dynamic path planning unit is used to perform dynamic path planning based on the current position and the target position to generate the optimal navigation path;

[0069] The command unit is used to generate control instructions based on the optimal navigation path and command the unmanned forklift to perform corresponding navigation and operations.

[0070] The navigation data acquisition unit includes:

[0071] A first acquisition module is used to acquire spatial distribution data of the indoor environment based on indoor sensors, and generate an indoor navigation topology network based on the spatial distribution data;

[0072] The second acquisition module is used to determine the driving area and obstacle distribution of the unmanned forklift in the indoor environment based on the navigation topology network and obtain real-time environmental perception data;

[0073] The third acquisition module is used to perform real-time mapping and positioning based on environmental perception data and use visual SLAM technology to determine the current precise position of the unmanned forklift and the preset target position.

[0074] Compared with the prior art, the present invention has the following advantages:

[0075] A method for indoor navigation of an unmanned forklift includes: obtaining navigation data from the unmanned forklift based on the indoor environment to determine the unmanned forklift's current and target locations; performing dynamic path planning based on the current and target locations to generate an optimal navigation path; and generating control instructions based on the optimal navigation path to direct the unmanned forklift to perform the corresponding navigation and operation. Through real-time path planning and dynamic adjustments, the unmanned forklift can quickly respond to environmental changes, reduce waiting time, and improve logistics efficiency.

[0076] Other features and advantages of the present invention will be set forth in the description which follows, and in part will be obvious from the description, or may be learned by practice of the present invention.

[0077] The technical solution of the present invention is further described in detail below through the accompanying drawings and embodiments. BRIEF DESCRIPTION OF THE DRAWINGS

[0078] The accompanying drawings are used to provide a further understanding of the present invention and constitute a part of the specification. Together with the embodiments of the present invention, they are used to explain the present invention and do not constitute a limitation of the present invention. In the accompanying drawings:

[0079] Figure 1 This is a flow chart of an indoor navigation system and navigation method for an unmanned forklift in an embodiment of the present invention;

[0080] Figure 2 A flowchart for determining the current position and target position of an unmanned forklift in an embodiment of the present invention;

[0081] Figure 3 This is a structural diagram of an indoor navigation system for an unmanned forklift in an embodiment of the present invention. DETAILED DESCRIPTION

[0082] The preferred embodiments of the present invention are described below with reference to the accompanying drawings. It should be understood that the preferred embodiments described herein are only used to illustrate and explain the present invention, and are not used to limit the present invention.

[0083] An embodiment of the present invention provides an indoor navigation method for an unmanned forklift, comprising:

[0084] S101: Obtain navigation data of the unmanned forklift based on the indoor environment to determine the current position and target position of the unmanned forklift;

[0085] S102: Perform dynamic path planning based on the current position and the target position to generate an optimal navigation path;

[0086] S103: Generate control instructions based on the optimal navigation path to instruct the unmanned forklift to perform corresponding navigation and operations.

[0087] The working principle of the above technical solution is as follows: the unmanned forklift perceives the indoor environment through the sensors it carries (such as lidar, depth camera, ultrasonic sensor) and builds map information of the surrounding environment. Lidar: measures the distance to the object by emitting laser and obtains two-dimensional or three-dimensional point cloud data of the environment. Depth camera: captures the depth information of the image and is used to identify objects and spatial structures. Ultrasonic sensor: uses sound wave ranging to detect short-range obstacles. The unmanned forklift updates its own position in real time through SLAM (Simultaneous Localization and Mapping) technology to determine its current position. The target position is usually set by the user through the scheduling system and described in the form of a coordinate system.

[0088] Based on the current position and target position, the system dynamically calculates the optimal navigation path through a path planning algorithm based on the current position and target position of the unmanned forklift and the environmental map.

[0089] A* (A-Star): A common heuristic path search algorithm that quickly finds the shortest path from a starting point to a destination. Dijkstra's algorithm: Ensures that all possible paths are traversed to find the optimal path, suitable for static or low-dynamic environments. RRT (Rapidly-Exploring Random Tree): Rapidly generates feasible paths in complex and dynamic environments.

[0090] Dynamic adjustment: If dynamic obstacles (such as other mobile devices or people) appear in the environment, the path planning algorithm will adjust the path in real time to avoid collision and ensure the continuity and safety of navigation.

[0091] Based on the optimal path, the system breaks it down into a series of control commands, including steering, acceleration, deceleration, and stopping. These commands are typically sent to the forklift's motion control system in the form of speed and angle commands. Upon receiving these commands, the unmanned forklift's drive system precisely executes the corresponding movements. Simultaneously, sensors monitor driving conditions in real time and transmit feedback back to the system for further navigation adjustments. For example, if the target location is 10 meters ahead of the forklift, the system generates a "go straight 10 meters" command. If an obstacle appears ahead, the system instantly generates a new "turn right to bypass the obstacle" command.

[0092] The beneficial effects of this technical solution include: through real-time path planning and dynamic adjustments, unmanned forklifts can quickly respond to environmental changes, reduce waiting time, and improve logistics efficiency. Leveraging sensors and dynamic path planning algorithms, unmanned forklifts can avoid obstacles in real time, effectively reducing the risk of collisions with other equipment or personnel. The generation of optimal paths reduces unnecessary detours, thereby reducing energy consumption and equipment wear. Unmanned forklifts can adapt to complex and changing indoor environments, maintaining a high level of autonomy and flexibility in dynamic environments, and supporting operational needs in a variety of scenarios.

[0093] In another embodiment, step S101 includes:

[0094] S1011: Acquire spatial distribution data of the indoor environment based on indoor sensors, and generate an indoor navigation topology network based on the spatial distribution data;

[0095] S1012: Based on the navigation topology network, determine the driving area and obstacle distribution of the unmanned forklift in the indoor environment and obtain real-time environmental perception data;

[0096] S1013: Based on environmental perception data, use visual SLAM technology to perform real-time mapping and positioning to determine the current precise position of the unmanned forklift and the preset target position.

[0097] Determining the current precise location of the unmanned forklift and the preset target location includes:

[0098] Based on the image in the environmental perception data, feature points in the image are extracted;

[0099] The system tracks the motion trajectory of the feature points and calculates the relative position changes of the feature points;

[0100] According to the relative position change, the motion trajectory of the unmanned forklift is derived;

[0101] Accumulate the spatial information of feature points to generate a two-dimensional or three-dimensional map of the environment;

[0102] Combine the generated environmental map with the sensor's odometry data to determine the precise current location of the unmanned forklift;

[0103] Get the preset target position;

[0104] Plan the navigation path of the unmanned forklift based on the current precise location and target location;

[0105] Determine the driving direction and speed of the unmanned forklift based on the navigation path;

[0106] If the unmanned forklift deviates from the navigation path, it adjusts the driving direction and speed to return to the navigation path;

[0107] Continuously acquire environmental perception data and repeat the above steps to achieve real-time mapping and positioning of the unmanned forklift;

[0108] The step of extracting feature points from the image includes:

[0109] Use image processing algorithms to identify salient features in images, such as corners, edges, or textures;

[0110] Generate descriptors for feature points to facilitate matching in consecutive image frames;

[0111] The system tracks the motion trajectory of the feature point including:

[0112] Match the same feature points between consecutively captured image frames;

[0113] Record the position changes of feature points in different image frames;

[0114] Derivation of the motion trajectory of the unmanned forklift includes:

[0115] Calculate the displacement and rotation of the unmanned forklift at each time point based on the position changes of the feature points;

[0116] Accumulate the displacement and rotation to obtain the overall motion trajectory of the unmanned forklift;

[0117] Generate a 2D or 3D map of your environment including:

[0118] Use the spatial distribution of feature points to construct the spatial structure of the environment;

[0119] Generate a 3D model of the environment (if a 3D map is selected) through a 3D reconstruction algorithm;

[0120] The combining of the environment map and the odometer data includes:

[0121] Fuse the map generated by visual SLAM with the odometry information provided by the sensor to improve positioning accuracy;

[0122] Optimizing the current position estimate of the unmanned forklift using a filtering algorithm (e.g., a Kalman filter);

[0123] Planning the navigation path of an unmanned forklift includes:

[0124] Marking the target location in an environment map;

[0125] Calculate the optimal path from the current precise location to the target location using a path planning algorithm (such as an A* algorithm);

[0126] Determining the driving direction and driving speed includes:

[0127] Adjust the driving direction of the unmanned forklift according to the curve and slope of the navigation path;

[0128] Set the corresponding driving speed according to the distance of the navigation path and the distribution of obstacles;

[0129] Adjusting the driving direction and speed includes:

[0130] Real-time monitoring of the driving status of unmanned forklifts;

[0131] According to the deviation, the driving parameters are dynamically adjusted to ensure that the unmanned forklift moves along the planned path;

[0132] Continuously acquiring the environmental perception data includes:

[0133] While driving, the unmanned forklift continuously uses the camera to capture images of the environment;

[0134] Update environmental maps and unmanned forklift positioning information in real time to adapt to the dynamically changing indoor environment.

[0135] The working principle of the above technical solution is as follows: To acquire spatial distribution data, the unmanned forklift uses indoor sensors (such as LiDAR and depth cameras) to scan the environment and collect indoor spatial distribution data. The LiDAR measures the distance between objects and the device, generating point cloud data (a three-dimensional spatial dataset). The depth camera obtains depth information from the image to supplement the object's height and outline data.

[0136] Topological network generation: Based on spatial distribution data, a navigational topological network for indoor environments is constructed. A navigational topological network is a structure based on nodes (locations) and edges (paths), describing the areas and paths a forklift can navigate. For example, in a warehouse, aisles between shelves are defined as edges, with each intersection or turn as a node.

[0137] The system stores the topological network as a graph data structure for subsequent path planning and navigation.

[0138] Driving area and obstacle distribution identification: Based on the navigation topology network, the system updates the forklift's driving area and obstacle distribution through real-time environmental perception data (such as the location of dynamic obstacles or new objects). Driving area: The system identifies the paths (such as aisles and open areas) that the forklift can traverse based on the navigation topology network. Obstacle detection: New obstacles (such as stacked goods or people) are detected using depth cameras or lidar, and the topology network is dynamically adjusted.

[0139] Environmental perception data is periodically scanned and updated, ensuring the automated forklift can quickly respond to environmental changes. For example, if the aisle ahead was previously clear, but sensors detect a pile of cargo, the system immediately marks the path as unavailable.

[0140] Visual SLAM technology provides real-time mapping and positioning. SLAM (Simultaneous Localization and Mapping) is a technology that simultaneously performs positioning (determining the device's own position) and mapping (generating a map of the environment). Visual SLAM uses camera image information to implement SLAM functionality. Compared to laser SLAM, visual SLAM is more suitable for texture-rich indoor environments.

[0141] Feature Extraction: Visual SLAM captures images through a camera and extracts feature points (such as corners, shelf edges, and landmark objects). Matching and Tracking: The system tracks these feature points, calculates their relative position changes, and derives the forklift's trajectory. Real-time Mapping: The spatial information of feature points is accumulated to generate a 2D or 3D map of the environment. Localization: The mapping results are combined with odometry data from sensors (such as displacement information from wheel encoders) to determine the forklift's current position.

[0142] For example, in a warehouse, an unmanned forklift uses visual SLAM to capture feature points along the shelf edges and continuously update its relative position. Based on a preset target location (such as shelf B2), the forklift can adjust its route in real time.

[0143] The beneficial effects of the above technical solution are as follows: Visual SLAM combined with environmental perception technology provides high-precision real-time positioning and mapping capabilities, ensuring the position accuracy of unmanned forklifts can reach the centimeter level. This is particularly important when operating in confined spaces (such as shelf aisles). Dynamic obstacle detection and real-time updates of the navigation topology network enhance the system's adaptability in complex environments, enabling forklifts to safely circumvent obstacles and select new optimal paths. Precise path planning based on the topology network reduces ineffective paths and idle time of forklifts, thereby improving logistics efficiency.

[0144] In another embodiment, step S102 includes:

[0145] S1021: Compare the current position of the unmanned forklift with the target position, calculate the shortest path, and optimize the path curve by combining the forklift's turning radius, acceleration, and deceleration parameters;

[0146] S1022: During the path optimization process, determine whether the path passes through a narrow passage or a complex area, obtain environmental data of the area, and make detailed adjustments to the path based on the size and driving capacity of the forklift;

[0147] S1023: Based on the driving characteristics of the forklift, determine the key turning points and parking points on the path, combine the environmental data, and make detailed adjustments to these key points to generate the optimal navigation path.

[0148] Among them, determining the key turning points and stopping points on the path includes:

[0149] Based on the forklift's driving characteristics data (driving characteristics data includes maximum turning angle, acceleration and inertia when parking), the turning points and parking points on the path are analyzed and marked as preliminary key turning points and parking points;

[0150] Based on the preliminary key turning points and stopping points, the positions of the turning points and stopping points are refined and adjusted based on environmental data such as obstacle distribution, lane width, and path slope, generating adjusted key turning points and stopping points;

[0151] Based on the adjusted key turning points and parking points, the forklift's driving route is planned, and the specific locations of each key turning point and parking point are clearly marked on the route.

[0152] The working principle of the above technical solution is as follows: the current position is compared with the target position, and the unmanned forklift obtains the current position in real time based on the aforementioned visual SLAM and sensor data. At this time, the unmanned forklift can determine the starting point and end point of the path planning by calculating the distance between the current position and the preset target position (for example, the target position may be in front of a specific shelf). Assume that the current position of the forklift is in the shelf A1 area of ​​the warehouse, and the target position is the shelf C3 area. The system first calculates the shortest path from A1 to C3, which can be achieved through the A* algorithm (a commonly used path planning algorithm).

[0153] Path optimization (taking into account turning radius, acceleration and deceleration), based on the calculation of the shortest path, further considers the physical characteristics of the forklift and optimizes the curve of the path. Turning radius: The minimum radius of the forklift when turning, usually determined by the length and wheelbase of the forklift. If there is a too sharp turn in the path, the forklift cannot pass smoothly, so the path needs to be adjusted so that the turning radius is not less than the minimum turning radius of the forklift. Acceleration and deceleration: The acceleration and braking capabilities of the forklift are limited. Therefore, the optimization of the path also needs to be based on the acceleration and deceleration of the forklift (for example: the maximum acceleration is 0.5m / s 2 , the maximum deceleration is -0.7m / s 2 ) to avoid sudden acceleration or sudden stops to ensure smooth and safe driving. If the path includes a sharp turn, the system will calculate the forklift's minimum turning radius (assuming it is 1.5 meters) and adjust the path so that the forklift does not hit obstacles when turning, while avoiding excessive acceleration or sudden braking.

[0154] To determine whether the path passes through narrow passages or complex areas, the system determines whether the path passes through narrow passages or complex areas based on the real-time position of the forklift and the preset path. Narrow passages: For example, the width of the passage between two shelves may be only 1.2 meters, while the width of the forklift is 1 meter, so this is a narrow passage. The system will automatically identify such passages and analyze them. Complex areas: For example, areas where goods are stacked in a messy manner, or areas where there are multiple obstacles (such as pallets, pedestrians, etc.). If there is a section in the path that needs to pass through a narrow passage between two shelves, the system will identify the difference between the width of the passage and the width of the forklift to determine whether the area is drivable.

[0155] Obtain environmental data and refine path adjustments. When the path passes through a narrow or complex area, the system will obtain detailed environmental data of the area through sensors. Based on the size of the forklift (such as width, length, height) and driving ability (such as turning radius), the system will make detailed adjustments to the path. For example, the path in a narrow area needs to be adjusted to a looser turning angle, or another route needs to be selected. If the width of the forklift is close to the width of the aisle when passing through a narrow aisle, the system will adjust the path to be slightly offset to prevent the side of the forklift from colliding with the shelf.

[0156] Identify key turning points and stopping points. In path planning, these are the points where the forklift must complete a turn or stop. The system uses the forklift's driving characteristics (such as maximum turning angle, acceleration, and inertia when stopped) to determine these key points on the path. For example, suppose there are several turns on the path, the forklift has a turning radius of 1.5 meters, and the angle of a turning point on the path is too large, exceeding the forklift's maximum turning angle. The system will adjust the path to ensure the forklift can complete the turn smoothly.

[0157] The system uses sensors to collect environmental data for detailed adjustments at each key point (e.g., presence of obstacles, aisle width, turning angles, etc.) and makes detailed adjustments to the route to ensure the forklift can safely pass through these key points. For example, if an obstacle (such as stacked goods) is present at a turning point, the system will replan the route to ensure the forklift can safely turn without hitting the obstacle.

[0158] The system generates an optimal navigation path. Ultimately, the system considers various factors (path optimization, narrow passages, turning points, and parking spots) to create the optimal navigation path. This path is not only the shortest possible, but also ensures the forklift reaches its destination smoothly, safely, and successfully. For example, after navigating complex areas, the path generated is no longer a simple straight line, but an optimized curved path. This takes into account factors such as the forklift's size, turning requirements, the complex environment, and turning radius, ensuring the entire path is as smooth and safe as possible.

[0159] The above technical solution offers the following benefits: By optimizing the forklift's path based on its physical characteristics (such as turning radius and acceleration), the system avoids collisions or unnecessary adjustments caused by improper path design. When the path passes through narrow passages or complex areas, the system automatically determines and refines the path to ensure safe passage for the forklift. Shortest path calculation and path curve optimization ensure that the forklift avoids wasted time during travel, reducing unnecessary detours or stops, thereby improving overall transportation efficiency. Real-time environmental data acquisition and path refinement prevent forklift stalls or accidents caused by poor path design in narrow or complex areas. Dynamic path adjustment: Path planning is not only static but also adapts in real time based on real-time environmental data (such as dynamic obstacles and passageway changes), ensuring the forklift's flexibility in dynamic environments. Taking into account the forklift's acceleration and deceleration limits, the system optimizes acceleration and braking points along the path, avoiding sudden acceleration or braking, resulting in smoother travel and reduced mechanical wear. Automated path calculation and optimization reduce the need for manual intervention and errors, while also improving system reliability and stability.

[0160] In another embodiment, step S103 includes:

[0161] S1031: Generate motion control instructions for the unmanned forklift based on the optimal navigation path. The motion control instructions include steering, acceleration, and deceleration.

[0162] S1032: driving the unmanned forklift along the planned path by executing control instructions;

[0163] S1033: Monitor the movement status of the unmanned forklift in real time and perform path deviation correction.

[0164] The working principle of the above technical solution is: the motion control instructions of the unmanned forklift are the core of its path planning and automatic driving. The motion control instructions are generated through the following steps:

[0165] Path Planning: The control system for an unmanned forklift relies on global navigation algorithms (such as the A* algorithm or the Dijkstra algorithm) and local obstacle avoidance algorithms (such as the dynamic window method). When an unmanned forklift needs to move from point A to point B in a warehouse, the path planning module comprehensively considers obstacle locations, the shortest path, and real-time dynamic information to generate an optimal path. The navigation path is represented by a series of coordinate points.

[0166] Motion command generation: Based on the coordinate points of the navigation path, the unmanned forklift's control system generates specific commands through the kinematic model, including:

[0167] Turn instructions: such as "turn left 15°" or "turn right 10°".

[0168] Acceleration command: such as "accelerate to 2m / s".

[0169] Deceleration command: such as "slow down to 0.5m / s".

[0170] For example: If an unmanned forklift detects a turn, the system will issue a command to "slow down to 1m / s and turn left 20°".

[0171] The execution phase is the process in which the unmanned forklift moves according to instructions: the system converts the instructions for steering, acceleration or deceleration into low-level control signals, and drives the forklift wheels to perform the action through the servo motor. For example, when receiving the instruction to "turn left 15°", the servo system adjusts the steering angle of the front wheels and slows down at the same time to ensure smooth steering. During driving, on-board sensors (such as odometers, lidars, and cameras) will provide real-time feedback on position information to ensure that the actual motion state of the unmanned forklift matches the planned path. For example, if the forklift deviates from the path by more than the set threshold, the system will send a new correction instruction, such as "turn right 3°" or "slightly accelerate."

[0172] Motion status monitoring: Multiple sensors are used to detect the speed, position, and angle of the unmanned forklift in real time. For example, when the LiDAR detects an obstacle, the system immediately triggers a deceleration command to avoid a collision. Path deviation correction: If the forklift deviates from the planned path, the system calculates a correction command based on the difference between the actual position and the target path. For example, if the forklift is intended to travel along a straight path but deviates due to uneven ground, the system will send a "turn left 2°" or "turn right 1°" command to adjust back to the correct path.

[0173] The benefits of this technical solution include: precise navigation path planning and motion control commands enable unmanned forklifts to efficiently complete complex tasks, such as quickly transporting goods in crowded warehouses. Real-time monitoring and deviation correction ensure safe operation in dynamic environments, preventing collisions with obstacles or other vehicles. When the LiDAR detects an obstacle within 0.5 meters, the unmanned forklift automatically slows down and stops to protect personnel. Real-time path deviation correction enables the unmanned forklift to respond to unexpected situations, such as slippery surfaces or temporary obstacles.

[0174] In another embodiment, the step S103 includes:

[0175] When an unmanned forklift is transporting goods, the status of the goods is monitored in real time based on the data from the unmanned forklift's indoor sensors. This includes detecting whether the goods are damaged or unbalanced.

[0176] Analyze cargo status data and identify potential cargo transportation risks in advance;

[0177] When abnormalities are detected in the cargo, the emergency braking mechanism is triggered to ensure the safety of the cargo and the unmanned forklift.

[0178] The working principle of the above technical solution is: the unmanned forklift uses onboard sensors to monitor the status of the cargo in real time, ensuring the safety and stability of the cargo during transportation. The following are the functions of several key sensors and the data processing process:

[0179] Acceleration sensor: used to monitor the vibration of the goods. By detecting whether the forklift has experienced violent acceleration, deceleration or sudden bumps during transportation, it can be determined whether the goods have been subjected to external force impact or imbalance. For example, if the forklift suddenly stops and the acceleration sensor data shows an abnormal impact, it means that the goods may have tilted or moved. Tilt sensor: used to detect the tilt angle of the goods. If the goods are tilted due to improper loading or other reasons, the tilt sensor will monitor the deviation from the normal horizontal plane in real time (such as a tilt of more than 10 degrees). Temperature and humidity sensor: When transporting goods that are easily affected by the environment (such as food, chemicals, etc.), the temperature and humidity sensor can monitor environmental parameters to prevent the goods from being affected by excessively high / low temperature or humidity during transportation. Visual sensor (such as a camera or a 3D scanner): used to monitor the appearance of the goods. Through real-time image processing and analysis, it can be detected whether the goods are damaged, such as whether the box is damaged or whether the packaging is intact.

[0180] The sensors collect various data (such as vibration acceleration, tilt angle, temperature and humidity) in real time and send them to the unmanned forklift's control system. The control system analyzes this data in real time. For example, if the acceleration sensor detects a vibration signal exceeding a set threshold, the control system will determine that the cargo has been subjected to a severe impact. If the tilt sensor detects that the cargo has deviated from its normal angle, the system will indicate that the cargo is unbalanced.

[0181] Cargo Tilt and Balance Analysis: The control system monitors the cargo's levelness in real time based on data from tilt sensors. For example, if cargo tilts beyond a certain angle on a forklift's pallet, the system identifies it as unbalanced, potentially causing it to fall or become damaged during transport. The system calculates whether the cargo presents a transportation risk based on historical data (such as the vibration tolerance and tilt angle of specific cargo types) and real-time sensor data. If the analysis indicates cargo instability, an alert is issued, notifying the operator to take appropriate action.

[0182] Cargo Damage Detection: Using visual sensors, the system regularly scans the cargo's exterior and processes the images. For example, by comparing images of the cargo before and after transport, the system can detect any damage, collapsed boxes, or loose packaging. If damage or packaging is detected, the system issues an alarm, alerting operators to adjust transportation methods or implement appropriate protective measures.

[0183] Environmental Risk Analysis: For environmentally sensitive goods (such as chemicals and cold chain products), data from temperature and humidity sensors can be used to determine whether the environment is suitable. If the temperature and humidity values ​​exceed the set range (such as temperature above 30°C or below -10°C, humidity above 80%, etc.), the system will issue a warning, prompting the operator to take measures (such as lowering the ambient temperature, opening windows for ventilation, etc.).

[0184] When the control system of the unmanned forklift detects that the cargo is in an abnormal state (such as tilting, damage or unsuitable environment), the system will automatically activate the emergency braking mechanism to ensure the safety of the cargo and the unmanned forklift.

[0185] Trigger conditions: The cargo is tilted too much: If the tilt angle of the cargo exceeds the set threshold (for example, more than 15 degrees), the system will determine that the cargo is unbalanced and trigger emergency braking to prevent the cargo from falling or further damage. Severe impact detection: If the acceleration sensor detects a large impact (such as sudden braking or collision with an obstacle), it is considered that the cargo may be damaged, and the system will immediately take emergency braking measures. Damaged cargo: If the visual sensor detects obvious damage on the surface of the cargo (such as broken outer packaging), the system will determine that there is an abnormality in transportation, trigger braking and remind the operator to check the status of the cargo. Environmental abnormality: If the temperature and humidity sensor detects that the transportation environment is beyond the safe range, the system will assess the risk and trigger the deceleration or braking mechanism to ensure the safety of the transportation process.

[0186] Emergency Braking Mechanism: Once an anomaly is detected, the control system immediately issues a braking command through the onboard control system (such as the electronic brake system), causing the forklift to slow down or come to a complete stop within a short period of time. By precisely controlling the braking force, secondary damage to the cargo caused by sudden braking is avoided. For example, if the tilt angle exceeds 15 degrees, the system will instruct the forklift to slow down and stop, ensuring that the cargo does not continue to tilt and fall or be damaged.

[0187] The beneficial effects of the above technical solution are: by monitoring the status of the goods in real time and triggering emergency braking when an abnormality occurs, it can effectively prevent transportation accidents caused by unbalanced goods, damage or unsuitable environments, and ensure the safety of the goods. The system can identify potential risks in advance before an abnormality occurs, such as tilted goods, damaged goods or transportation environment problems, so as to take preventive measures (such as adjusting the placement of goods or adjusting transportation conditions) to reduce the probability of accidents. Based on sensor data and intelligent analysis, unmanned forklifts can independently judge the status of goods and automatically execute control decisions without relying on human intervention. This automation improves overall transportation efficiency and reduces the workload of operators.

[0188] In another embodiment, determining a driving area and obstacle distribution of an unmanned forklift in an indoor environment includes:

[0189] When the obstacle density detected in the driving area by real-time environmental perception data exceeds the preset threshold, it is determined that there is a potential obstacle risk in the current driving area, a red mark is added to the navigation topology network, and an obstacle warning notification is sent to the warehouse management system;

[0190] Otherwise, the current driving area is determined to be in normal operation, and a green mark is added to the navigation topology network;

[0191] If there are potential obstacle risks in the current driving area, including:

[0192] Obtain the real-time navigation signal of the unmanned forklift, determine the destination of the unmanned forklift based on the navigation information, and generate a recommended driving route corresponding to the destination based on the navigation topology network and real-time environmental perception data;

[0193] Obtain the expected obstacle density for each time period corresponding to each recommended route, determine whether the expected obstacle density for the time period corresponding to the predicted travel time of the undetermined driving area exceeds the preset density of the recommended driving area, and if so, regenerate the recommended driving route;

[0194] Otherwise, the recommended driving route is sent to the corresponding area of ​​the navigation display interface of the unmanned forklift for display;

[0195] After confirming that the unmanned forklift has adopted the recommended driving route, based on the current driving data of the unmanned forklift, the time period corresponding to the unmanned forklift's arrival at each section of the recommended route and its corresponding driving area is predicted to generate the recommended driving route;

[0196] Based on the recommended driving route, the expected obstacle density of each recommended driving area is marked on the navigation topology network.

[0197] The working principle of the above technical solution is as follows: Obstacle density: refers to the number of obstacles detected in a unit area, which can be calculated through data collected by environmental perception equipment such as laser radar (LiDAR) and cameras. For example, if there are 10 obstacles within 5 meters in a certain area, the obstacle density is 2 per square meter. Preset threshold: A standard value set according to the driving needs of the system. If the obstacle density exceeds 1 per square meter, it is judged as a high-density area. The real-time environmental perception system compares the obstacle density data with the preset threshold: Above the threshold: It is judged that there is a potential obstacle risk, and the system adds a red mark to the corresponding area on the navigation topology network, and sends an early warning notification to the warehouse management system. Below the threshold: It is judged to be in normal operating state, and the system adds a green mark to the navigation topology network.

[0198] Navigation topology network: A virtualized navigation map that identifies all drivable paths, nodes, and areas within a warehouse environment. For example, a complex warehouse may include multiple paths from Area A to Area B. Nodes between these paths may mark turning points, intersections, and other locations. The system acquires real-time navigation signals from unmanned forklifts to determine their current destinations. Based on the navigation topology network and real-time environmental perception data, it generates one or more recommended routes to the destination.

[0199] Expected Obstacle Density: Combining historical and real-time data, the system predicts the obstacle density of a specific area within a specific timeframe. For example, based on a statistical model, the obstacle density of a specific route is expected to reach 1.2 obstacles per square meter at 9:00 AM. The system determines whether the proposed travel area exceeds the preset density threshold for the recommended travel area during the period corresponding to the predicted travel time. If the threshold is exceeded, the recommended route is regenerated, avoiding high-density areas. If the threshold is not exceeded, the recommended route is confirmed.

[0200] The recommended route is provided to the unmanned forklift through the navigation display interface. After the unmanned forklift confirms the recommended route, the system predicts the specific time period for the forklift to arrive at each section of the route. Expected obstacle density marking: The predicted obstacle density results for each section and time period are displayed on the navigation topology network for subsequent scheduling reference.

[0201] The beneficial effects of the above technical solution are as follows: By monitoring obstacle density, the system can identify potential risk areas in advance and reduce the possibility of unmanned forklifts stalling or colliding in high-risk areas. Real-time early warning notifications allow warehouse managers to quickly adjust scheduling and workflows. Based on real-time environmental perception and predicted obstacle density data, the system can dynamically adjust recommended routes to avoid congestion and risks and improve transportation efficiency. Expected obstacle density markings provide data support for path planning, making recommended routes more accurate. This reduces the time unmanned forklifts pause or detour due to excessive obstacle density, thereby improving the overall task completion speed. Automated navigation and real-time optimization mechanisms reduce the need for manual intervention. The navigation topology network combines real-time and historical data to construct a dynamically updated intelligent warehouse map, making the entire system highly adaptable. Density marking and path optimization functions can help managers develop more accurate warehouse scheduling strategies and improve warehouse operational efficiency.

[0202] In another embodiment, determining whether the path passes through a narrow passage or a complex area includes:

[0203] Obtain path planning data and build a path optimization model;

[0204] The path optimization model is configured with a path detection window set, and each path detection window in the path detection window set is provided with a path complexity evaluation image;

[0205] Collecting path travel data, dividing the path travel data into path detection windows, and inputting the data into each path detection window for window filling;

[0206] Obtaining the window filling result and determining the path complexity based on the ratio of the filling area of ​​the path detection window to the total area of ​​the path detection window in the path optimization model;

[0207] The path optimization model includes a path detection window and a path evaluation mechanism;

[0208] The path detection window is used to import the path planning data into the corresponding path detection window according to the corresponding path segment;

[0209] The path evaluation mechanism is embedded in the path detection window and is used to compare and cover the environmental data imported into the path detection window and output the coverage area;

[0210] Route adjustment includes: obtaining route travel data;

[0211] Input the path driving data into the preset generative adversarial network for path division, and obtain the path identification output by the generative adversarial network;

[0212] The generative adversarial network consists of a generator and a dual-output discriminator;

[0213] Generative adversarial networks are trained as follows:

[0214] Obtain path planning samples, path type parameters and path import identifier loading parameters;

[0215] Inputting the path planning sample into the generator for pre-training to obtain a first recognition network output by the generator;

[0216] The path planning sample, path type parameter and path import identifier are loaded into the discriminator for training to obtain the discriminant branch of the discriminator;

[0217] Determine the path corresponding to the path planning data based on the discriminant branch and the first identification network, and generate a path identifier for importing the path data into the identification mechanism;

[0218] According to the path identification, the path planning data is imported into different path detection windows for path filling;

[0219] Also includes:

[0220] Get the filling data of each path detection window;

[0221] Predetermining the upper and lower thresholds of the path complexity corresponding to each path detection window;

[0222] Based on the upper and lower thresholds, the path data is monitored for over-limit, and the abnormal complexity is determined based on the over-limit monitoring results;

[0223] Perform anomaly evaluation on the path data according to the anomaly complexity of each path detection window to determine the path evaluation data;

[0224] Each path is automatically refined and adjusted based on the path evaluation data.

[0225] The above technical solution works as follows: The path optimization model is a system tool used to optimize navigation paths. It includes a set of multiple path detection windows for autonomous forklift navigation, each responsible for monitoring and assessing the complexity of a specific path segment. For example, when an autonomous forklift needs to navigate between warehouse shelves, these windows can help identify narrow aisles or obstacles.

[0226] In indoor navigation scenarios for unmanned forklifts, the forklift must plan the optimal path within a complex environment to improve transport efficiency and avoid collisions with obstacles. The path optimization model is built based on the path planning data and includes a set of path detection windows and a path evaluation mechanism.

[0227] The path detection window is a virtual framework that divides the path environment and is used to analyze the complexity of different areas. For example, the path complexity of a forklift is low in a spacious aisle area, but the complexity increases in the narrow alleys between high shelves. Each window uses a path complexity evaluation image to evaluate the characteristics of the forklift's driving path, such as aisle width, obstacle distribution, etc. Path detection window collection: The navigation environment is divided into several small areas, each of which is called a path detection window. For example, narrow areas between shelves and wide open areas in a warehouse can be divided by detection windows. Path complexity evaluation image: Each detection window generates a specific complexity evaluation image based on the data collected from the environment to reflect the complexity of the area, such as the density of obstacles or the width of the space.

[0228] While performing tasks, unmanned forklifts collect real-time driving data (such as location, speed, and LiDAR point cloud data). This data is divided according to the path detection window and entered into the window for filling. Window filling: Driving data is gradually accumulated and filled into the detection window to form a model of the actual driving area in the corresponding window. The ratio of the filled area to the total area of ​​the path detection window can be used to reflect the complexity of the path.

[0229] Through the path optimization model, the ratio of the filled area to the total area of ​​the detection window is analyzed. The more complex the window, the lower the filling ratio, which means a narrow channel or complex area.

[0230] Generative adversarial networks (GANs) are used to further optimize path identification. GAN consists of a generator and a dual-output discriminator: Generator: Generates a preliminary path planning plan after pre-training and outputs the first recognition network. Discriminator: Combines path type parameters and environmental loading identifiers to determine whether the generated path identification meets the actual planning requirements. In unmanned forklift navigation, GAN can identify different path types (such as spacious areas, narrow passages, or complex intersections). After the generative adversarial network is trained, it can generate path identification based on path planning data, which is used to automatically match the path detection window and complete the path filling.

[0231] Each path detection window has upper and lower complexity thresholds (e.g., the width of a space suitable for a forklift). When the filling result exceeds the threshold, the system flags the complexity as abnormal, triggering the evaluation and path optimization mechanism. By integrating the abnormal evaluation results of the path detection windows, the system automatically optimizes the path planning scheme. For example, it can adjust the forklift path to narrow areas or recommend a new path that avoids high-complexity areas.

[0232] For example: the narrow aisle between warehouse shelves is divided into a path detection window, which generates a complexity evaluation image. The image shows the location of the shelves and the width of the passable area. The complexity evaluation reflects the difficulty of the forklift passing through. When the unmanned forklift drives into the aisle, the system collects its real-time position and lidar data and fills it into the corresponding detection window. If the ratio of the filled area to the total area is low, it means that the complexity of this area is high (such as the presence of obstacles or narrow space). Generative adversarial network path optimization: GAN is used to refine and identify the path. For example, the generator recommends a path to bypass obstacles, and the discriminator determines whether the recommended path meets the actual capabilities of the forklift (such as the turning radius). Over-limit monitoring and anomaly assessment: If the complexity evaluation result exceeds the threshold (for example, the width of the aisle between the shelves is not enough to accommodate the forklift), the system will mark the abnormal complexity and re-plan the detour path. The beneficial effect of the above technical solution is: through the path optimization model, complex paths can be automatically refined into low-complexity paths, thereby reducing human intervention and improving the navigation efficiency of unmanned forklifts. The path detection window set refines the path segment assessment, allowing the forklift to accurately perceive environmental changes. For example, in a dynamic environment, the system can adjust the path in a timely manner to avoid collisions. The introduction of a generative adversarial network enables accurate identification of path types. Especially in high-risk areas (such as narrow alleys or intersections), path markers can quickly prompt the forklift to adopt a slowdown or detour strategy. The path complexity assessment mechanism enables the system to adapt to changing indoor environments, whether in spacious areas or crowded aisles, and can make intelligent decisions.

[0233] In another embodiment, real-time monitoring of cargo status includes:

[0234] Sensors on unmanned forklifts collect real-time dynamic data on cargo, including displacement, vibration, tilt angle, and other physical characteristics related to cargo status.

[0235] Process and analyze the collected dynamic data to identify the movement state of the goods during transportation and detect whether there is abnormal vibration or tilt;

[0236] If abnormal vibration or tilt is detected, the status of the cargo is determined to be damaged or unbalanced. If so, an early warning message is generated and a processing flow is triggered;

[0237] Through the cargo status assessment model, a comprehensive assessment of the cargo status is conducted based on sensor data to determine whether the cargo is in normal transportation status. If the cargo status is abnormal, corrective measures are taken according to preset rules;

[0238] The cargo status assessment results are combined with real-time monitoring data to generate a cargo status report, and the transportation path or movement of the unmanned forklift is adjusted in real time to avoid cargo damage or imbalance problems.

[0239] The working principle of this technical solution is that unmanned forklifts use sensors installed to monitor the status of cargo during transportation in real time. For example, accelerometers track cargo displacement, tilt sensors detect the cargo's tilt angle, and vibration sensors identify vibrations during transportation. These sensors continuously collect data and provide feedback to the system.

[0240] The collected dynamic data enters the processing system for preliminary analysis. The system determines the cargo's motion state by comparing sensor data under normal transport conditions (such as stable displacement and slight vibration) with the actual collected data. For example, if the cargo's tilt angle exceeds a preset safety threshold or the vibration amplitude increases abnormally, the system will determine that an abnormality exists and further analyze whether the cargo is damaged or unbalanced.

[0241] By further processing and performing pattern recognition on the data, the system can identify any anomalies that may occur during cargo transportation. If the cargo experiences sudden, severe vibration or tilt during transport, the system will issue an early warning signal, indicating potential damage or imbalance. Comprehensive analysis of sensor data plays a key role in this process.

[0242] During data analysis, the system uses a cargo condition assessment model to determine the overall condition of the cargo. This model can be considered a set of algorithms based on sensor data. It combines various physical characteristics of the cargo during transportation (such as tilt, vibration, and position changes) to determine whether the cargo is in a safe and stable state. If the model determines that the cargo is tilted too much or vibrating abnormally, it will determine that the cargo is in an abnormal state and issue an alarm.

[0243] When the cargo status assessment model determines that the cargo is in an abnormal state, the system automatically generates an alert and triggers the appropriate processing flow. This may include rerouting the forklift, adjusting the cargo's position, or taking other corrective measures to avoid cargo damage or transport imbalance.

[0244] If an anomaly is detected, the system combines real-time monitoring data with cargo status assessment results to generate a detailed cargo status report and adjust the forklift's path or actions based on the assessment results. For example, if the system detects that the cargo is tilted at a corner, it can instruct the forklift to adjust its speed or route to avoid damage to the cargo due to excessive movement.

[0245] The beneficial effects of this technical solution are as follows: By monitoring cargo dynamic data in real time, the system can promptly detect anomalies during transportation, such as tilting or excessive vibration, thereby preventing losses caused by imbalance or damage during transportation. This early warning and response process significantly improves cargo transportation safety. The cargo status assessment model comprehensively analyzes data from multiple sensors to provide more accurate assessments of cargo transportation status. It not only determines in real time whether cargo is in normal transportation status, but also provides clear status assessment reports in abnormal situations, helping operators take timely measures.

[0246] In another embodiment, an indoor navigation system for an unmanned forklift includes:

[0247] A navigation data acquisition unit, used to acquire navigation data of the unmanned forklift based on the indoor environment and determine the current position and target position of the unmanned forklift;

[0248] A dynamic path planning unit is used to perform dynamic path planning based on the current position and the target position to generate the optimal navigation path;

[0249] The command unit is used to generate control instructions based on the optimal navigation path and command the unmanned forklift to perform corresponding navigation and operations.

[0250] Generate optimal navigation paths, including:

[0251] Compare the current position of the unmanned forklift with the target position to calculate the shortest path. At the same time, the path curve is optimized by combining the forklift's turning radius, acceleration, and deceleration parameters.

[0252] During the path optimization process, the system determines whether the path passes through narrow passages or complex areas, obtains environmental data of the area, and makes detailed adjustments to the path based on the size and driving capacity of the forklift.

[0253] Based on the driving characteristics of the forklift, the key turning points and parking points on the path are determined. Combined with environmental data, these key points are refined and adjusted to generate the optimal navigation path.

[0254] Command unmanned forklifts to perform corresponding navigation and operations, including:

[0255] Generate motion control instructions for the unmanned forklift based on the optimal navigation path, including steering, acceleration, and deceleration;

[0256] By executing control instructions, the unmanned forklift is driven along the planned path;

[0257] Monitor the movement status of unmanned forklifts in real time and correct path deviations.

[0258] After instructing the unmanned forklift to perform the corresponding navigation and operation steps, including:

[0259] When an unmanned forklift is transporting goods, the status of the goods is monitored in real time based on the data from the unmanned forklift's indoor sensors. This includes detecting whether the goods are damaged or unbalanced.

[0260] Analyze cargo status data and identify potential cargo transportation risks in advance;

[0261] When abnormalities are detected in the cargo, the emergency braking mechanism is triggered to ensure the safety of the cargo and the unmanned forklift.

[0262] The working principle of this technical solution is as follows: The navigation data acquisition unit uses sensor technologies such as lidar, visual cameras, inertial navigation systems (INS), RFID tags, or ultra-wideband (UWB) to scan the indoor environment in real time and collect environmental characteristic data. Using the SLAM (Simultaneous Localization and Mapping) algorithm, the sensor data is used to construct an environmental map and simultaneously determine the current position of the unmanned forklift. The user determines the desired location of the unmanned forklift by inputting target coordinates or providing target location data through a warehouse management system (WMS).

[0263] The dynamic path planning unit receives real-time environmental maps and forklift location information based on the current and target locations provided by the navigation data acquisition unit. This unit uses the A* algorithm, Dijkstra algorithm, RRT (Rapidly Exploring Random Trees), or an intelligent algorithm based on reinforcement learning to analyze the distribution of indoor obstacles and traffic space, dynamically planning an optimal navigation path. For dynamic obstacles encountered by the unmanned forklift during operation (such as other forklifts or pedestrians), the system will replan the path in real time to ensure that the forklift can reach the target location safely and quickly.

[0264] Based on the optimal path provided by the dynamic path planning unit, the command unit generates a series of control commands, such as forward movement, steering, stopping, and obstacle avoidance. These commands are transmitted via the motion controller to the unmanned forklift's drive system, steering system, and sensor module, enabling the truck to precisely follow the planned path and perform tasks such as loading and unloading. During this process, the command unit continuously receives feedback from the navigation data acquisition unit, monitoring the truck's position and status in real time to ensure successful mission completion.

[0265] The beneficial effects of this technical solution include: dynamic path planning enables real-time selection of optimal routes, reducing idle time and route redundancy for forklifts, and improving navigation efficiency. Automated command and operation eliminate the uncertainty inherent in manual operation, accelerating task execution. The navigation data acquisition unit utilizes high-precision sensors and SLAM technology, enabling unmanned forklifts to achieve high-precision positioning and path tracking in complex indoor environments. Feedback mechanisms and dynamic adjustment capabilities ensure that navigation paths remain valid and minimize deviations.

[0266] In another embodiment, the navigation data acquisition unit includes:

[0267] A first acquisition module is used to acquire spatial distribution data of the indoor environment based on indoor sensors, and generate an indoor navigation topology network based on the spatial distribution data;

[0268] The second acquisition module is used to determine the driving area and obstacle distribution of the unmanned forklift in the indoor environment based on the navigation topology network and obtain real-time environmental perception data;

[0269] The third acquisition module is used to perform real-time mapping and positioning based on environmental perception data and use visual SLAM technology to determine the current precise position of the unmanned forklift and the preset target position.

[0270] The working principle of the above technical solution is as follows: The first acquisition module (acquiring indoor spatial distribution data) uses a series of indoor sensors (such as lidar, infrared sensors, ultrasonic sensors, UWB, and visual cameras) to scan the surrounding environment in real time, collecting spatial distribution information including walls, doorways, aisles, shelves, etc. Based on different measurement principles, the sensors generate 3D point cloud data or 2D plane data of the environment. By analyzing the data collected by the sensors and using spatial analysis algorithms (such as clustering algorithms, geometric modeling, and graph theory algorithms), a navigation topology network of the indoor environment is constructed. The topology network is based on the connectivity of important elements in the environment (such as aisles, nodes, intersections, and obstacles), forming a network diagram that ensures that unmanned forklifts can plan their routes based on this network information. Through this topology construction, unmanned forklifts can clearly identify each traversable area and the connectivity between them, providing a foundation for subsequent route planning and navigation. The topology network not only reflects spatial distribution information but also can dynamically adjust to adapt to changes in the indoor environment (such as shelf locations and temporary obstacles). The second acquisition module (environmental perception and driving area identification) builds on the navigation topology network generated by the first module and further uses sensors to acquire data on obstacles, driving areas, and other aspects of the environment. Obstacle and driving area identification: Using lidar or visual sensors, the system detects dynamic obstacles (such as other equipment, personnel, or temporary obstacles) and static obstacles (such as walls and fixed shelves) in the environment in real time. By integrating with the topological network, it determines which areas are passable and which areas are occupied by obstacles, further refining the forklift's driving area. The combination of real-time environmental perception data and the navigation topological network enables the system to promptly update the obstacle distribution when the environment changes, automatically adjusting the unmanned forklift's driving path or designated area.

[0271] The third acquisition module (visual SLAM and positioning) is based on real-time environmental perception data. The third acquisition module uses visual sensors (such as cameras, depth cameras, etc.) combined with SLAM (simultaneous localization and mapping) technology to perform real-time map construction and positioning. Visual SLAM technology converts image data in the environment into map information through feature point matching, image processing and optimization algorithms, and at the same time identifies the current position of the forklift. The SLAM system updates the map of the indoor environment in real time by continuously scanning and matching image features, and combines the inertial navigation system (INS) and other sensor data to accurately calculate the current position of the unmanned forklift. At the same time, the preset target position is calibrated in the map to provide an accurate basis for the forklift's path planning. By continuously comparing the real-time acquired data with the mapping results, visual SLAM can ensure high-precision positioning of the forklift in complex environments, and can reliably navigate even in indoor environments without GPS signals.

[0272] The beneficial effects of the above technical solution are as follows: The navigation topology network generated by the first acquisition module enables the unmanned forklift to accurately understand the spatial distribution and traffic conditions of the indoor environment, providing effective support for subsequent path planning and behavioral decision-making. The environmental perception data acquired in real time by the second acquisition module can not only identify static obstacles, but also dynamically detect and respond to temporary obstacles such as people, equipment, or sudden obstacles, greatly enhancing the flexibility and adaptability of the system. The third acquisition module achieves precise positioning through visual SLAM technology, ensuring that the unmanned forklift can accurately locate its position in complex indoor environments, unaffected by environmental changes and variable factors.

[0273] Obviously, those skilled in the art may make various changes and modifications to the present invention without departing from the spirit and scope of the present invention. Thus, if such changes and modifications fall within the scope of the claims and their equivalents, the present invention is intended to include such changes and modifications.

Claims

1. An indoor navigation method for an unmanned forklift, characterized in that: include: S101: Obtain navigation data of the unmanned forklift based on the indoor environment to determine the current position and target position of the unmanned forklift; S102: Perform dynamic path planning based on the current position and the target position to generate an optimal navigation path; S103: Generate control instructions based on the optimal navigation path to instruct the unmanned forklift to perform corresponding navigation and operations.

2. The method for indoor navigation of an unmanned forklift according to claim 1, characterized in that: Step S101 includes: S1011: Acquire spatial distribution data of the indoor environment based on indoor sensors, and generate an indoor navigation topology network according to the spatial distribution data; S1012: Based on the navigation topology network, determine the driving area and obstacle distribution of the unmanned forklift in the indoor environment, and obtain real-time environmental perception data; S1013: Based on the environmental perception data, use the visual SLAM technology to perform real-time mapping and positioning to determine the current precise position of the unmanned forklift and the preset target position.

3. The indoor navigation method of an unmanned forklift according to claim 1, characterized in that: Step S102 includes: S1021: Compare the current position of the unmanned forklift with the target position, calculate the shortest path, and optimize the path curve by combining the turning radius, acceleration and deceleration parameters of the forklift; S1022: During the path optimization process, determine whether the path passes through a narrow passage or a complex area, obtain environmental data of the area, and make detailed adjustments to the path according to the size and driving capacity of the forklift; S1023: Based on the driving characteristics of the forklift, determine the key turning points and parking points on the path, combine the environmental data, refine and adjust these key points, and generate the optimal navigation path.

4. The indoor navigation method of an unmanned forklift according to claim 1, characterized in that: Step S103 includes: S1031: Generate motion control instructions for the unmanned forklift according to the optimal navigation path, the motion control instructions including steering, acceleration and deceleration; S1032: driving the unmanned forklift along the planned path by executing the control command; S1033: Monitor the movement status of the unmanned forklift in real time and perform path deviation correction.

5. The method for indoor navigation of an unmanned forklift according to claim 1, characterized in that: The following steps are included after step S103: When the unmanned forklift is transporting goods, the status of the goods is monitored in real time based on the indoor sensor data of the unmanned forklift. The status of the goods includes detecting whether the goods are damaged or unbalanced. Analyze cargo status data and identify potential cargo transportation risks in advance; When abnormalities in the cargo are detected, the emergency braking mechanism is triggered to ensure the safety of the cargo and the unmanned forklift.

6. The method for indoor navigation of an unmanned forklift according to claim 2, characterized in that: Determine the driving area and obstacle distribution of the unmanned forklift in the indoor environment, including: When the obstacle density in the driving area detected in the real-time environmental perception data exceeds the preset threshold, it is determined that there is a potential obstacle risk in the current driving area, a red mark is added to the navigation topology network, and an obstacle warning notification is sent to the warehouse management system; Otherwise, it is determined that the current driving area is in a normal operating state, and a green mark is added to the navigation topology network; If there are potential obstacle risks in the current driving area, including: Obtain the real-time navigation signal of the unmanned forklift, determine the destination of the unmanned forklift based on the navigation information, and generate a recommended driving route corresponding to the destination based on the navigation topology network and real-time environmental perception data; Obtain the expected obstacle density for each time period corresponding to each recommended route, determine whether the expected obstacle density for the time period corresponding to the predicted travel time of the pending driving area exceeds the preset density of the recommended driving area, and if so, regenerate the recommended driving route; Otherwise, the recommended driving route is sent to the corresponding area of ​​the navigation display interface of the unmanned forklift for display; After confirming that the unmanned forklift adopts the recommended driving route, based on the current driving data of the unmanned forklift, the time period corresponding to the unmanned forklift arriving at each section of the recommended route and its corresponding driving area is predicted to generate the recommended driving route; Based on the recommended driving route, the expected obstacle density of each recommended driving area is marked on the navigation topology network.

7. The method for indoor navigation of an unmanned forklift according to claim 3, characterized in that: Determine whether the path passes through narrow passages or complex areas, including: Obtain path planning data and build a path optimization model; The path optimization model is configured with a path detection window set, and each path detection window in the path detection window set is provided with a path complexity evaluation image; Collecting path travel data, dividing the path travel data according to path detection windows, and inputting each path detection window for window filling; Obtaining a window filling result, and determining the path complexity based on a ratio of a filling area of ​​the path detection window to a total area of ​​the path detection window in a path optimization model; The path optimization model includes a path detection window and a path evaluation mechanism; The path detection window is used to import the path planning data into the corresponding path detection window according to the corresponding path segment; The path evaluation mechanism is embedded in the path detection window to compare and cover the environmental data imported into the path detection window and output the coverage area; The route adjustment includes: obtaining route travel data; Input the path driving data into the preset generative adversarial network for path division, and obtain the path identification output by the generative adversarial network; According to the path identification, the path planning data is imported into different path detection windows for path filling.

8. The method for indoor navigation of an unmanned forklift according to claim 5, characterized in that: Real-time monitoring of cargo status, including: The sensors on the unmanned forklift collect the dynamic data of the cargo in real time. The dynamic data includes the displacement, vibration, tilt angle and physical characteristics related to the cargo status. Process and analyze the collected dynamic data to identify the movement state of the goods during transportation and detect whether there is abnormal vibration or tilt; If abnormal vibration or tilt is detected, determine whether the cargo is damaged or unbalanced. If so, generate an early warning message and trigger the processing flow; Through the cargo status assessment model, the cargo status is comprehensively assessed based on sensor data to determine whether the cargo is in a normal transportation state. If the cargo status is abnormal, corrective measures are taken according to preset rules; The cargo status assessment results are combined with real-time monitoring data to generate a cargo status report, and the transportation path or action of the unmanned forklift is adjusted in real time to avoid cargo damage or imbalance problems.

9. An indoor navigation system for an unmanned forklift, characterized in that: include: A navigation data acquisition unit, used to acquire navigation data of the unmanned forklift based on the indoor environment, and determine the current position and target position of the unmanned forklift; A dynamic path planning unit is used to perform dynamic path planning based on the current position and the target position to generate an optimal navigation path; The command unit is used to generate control instructions based on the optimal navigation path and command the unmanned forklift to perform corresponding navigation and operations.

10. The unmanned forklift indoor navigation system according to claim 9, characterized in that: The navigation data acquisition unit includes: A first acquisition module is used to acquire spatial distribution data of an indoor environment based on indoor sensors, and generate an indoor navigation topology network according to the spatial distribution data; The second acquisition module is used to determine the driving area and obstacle distribution of the unmanned forklift in the indoor environment based on the navigation topology network, and obtain real-time environmental perception data; The third acquisition module is used to perform real-time mapping and positioning based on environmental perception data and use visual SLAM technology to determine the current precise position and preset target position of the unmanned forklift.

Images