Storage robot intelligent obstacle avoidance system under Internet of Things

Abstract

The invention discloses a storage robot intelligent obstacle avoidance system under the Internet of Things, and belongs to the technical field of the Internet of Things, and the system comprises a multi-source data collection and fusion module, a path planning module, a multi-robot cooperation module, an execution control module, a task scheduling module, a safety monitoring module and a self-optimization module. According to the method, laser radar and visual data are integrated into a dynamic occupation grid map, a dynamic obstacle trajectory prediction model is introduced, active avoidance of moving obstacles is realized, and path smoothness and energy consumption efficiency can be optimized while a safe distance is ensured by quantifying distance and time dual cost and path planning. Besides, the path is decomposed into a time interval attitude sequence, and a curvature optimization objective function is combined, so that the robot can still keep efficient and stable obstacle avoidance capability in a storage environment with dense people streams or frequent movement, and the task execution success rate is remarkably improved.

Description

Technical Field

[0001] This invention belongs to the field of Internet of Things (IoT) technology, specifically referring to an intelligent obstacle avoidance system for warehouse robots under the Internet of Things. Background Technology

[0002] With the development of the Internet of Things and automation technologies, warehouse robots are widely used in logistics sorting, cargo handling, and other scenarios. However, existing obstacle avoidance systems for warehouse robots still face many technical bottlenecks.

[0003] However, existing obstacle avoidance methods for warehouse robots still have certain shortcomings. Existing methods, which rely on static grid maps or single obstacle avoidance algorithms, are unable to cope with the real-time trajectory changes of dynamic obstacles. They only calculate obstacle risks based on distance costs and do not comprehensively consider the movement speed and time urgency of dynamic obstacles. The generated paths may have abrupt changes, resulting in unstable robot movement, increased motor wear and energy consumption. They rely on centralized scheduling, which is prone to task blockage due to communication delays or single points of failure. They do not dynamically adjust robot priorities, causing high-priority tasks to be delayed due to low-priority robots occupying the path. When conflicts are detected, they are often resolved through global path replanning, but without combining local dynamic adjustment strategies, resulting in wasted computing resources and decreased path efficiency. To address these issues, an intelligent obstacle avoidance system for warehouse robots under the Internet of Things is proposed. Summary of the Invention

[0004] The purpose of this invention is to provide an intelligent obstacle avoidance system for warehouse robots under the Internet of Things, so as to solve the problems mentioned in the background art.

[0005] To achieve the above objectives, the present invention provides the following technical solution: an intelligent obstacle avoidance system for warehouse robots under the Internet of Things, comprising a multi-source data acquisition and fusion module, a path planning module, a multi-robot collaboration module, an execution control module, a task scheduling module, a safety monitoring module, and a self-optimization module;

[0006] The multi-source data acquisition and fusion module integrates environmental obstacle information based on the raw data from lidar and visual camera sensors through a multi-sensor adaptive fusion algorithm.

[0007] The path planning module generates obstacle avoidance paths based on obstacle location and dynamic trajectory data provided by the multi-source data acquisition and fusion module, combined with the target assigned by the task scheduling module, using a hybrid path planning algorithm.

[0008] The multi-robot collaboration module performs path conflict detection and collaborative adjustment based on the path information from the path planning module and the real-time status shared with other robots through a distributed communication protocol.

[0009] The execution control module controls the robot's motion posture through a PID controller and a differential drive system, based on the obstacle avoidance path instructions output by the path planning module.

[0010] The task scheduling module allocates robot task targets based on the task requirements issued by the warehouse management system and the path information generated by the path planning module through a dynamic allocation algorithm.

[0011] The security monitoring module, based on the sensor data from the multi-source data acquisition and fusion module and the feedback status from the execution control module, uses an anomaly detection mechanism to provide early warnings of hardware failures and environmental risks, and triggers strategy iterations of the self-optimization module when an anomaly occurs.

[0012] The self-optimization module iteratively updates the obstacle avoidance strategy and path planning algorithm based on the abnormal feedback from the safety monitoring module and the historical obstacle avoidance data from the path planning module through a reinforcement learning model.

[0013] Preferably, the multi-source data acquisition and fusion module integrates environmental obstacle information based on the raw data from the lidar and visual camera sensors through a multi-sensor adaptive fusion algorithm; by selecting lidar and visual camera as core data sources, configuring the communication protocols and data formats of each sensor, and setting the sampling frequency, resolution, and timestamp synchronization mechanism of the sensors.

[0014] Preferably, the multi-source data acquisition and fusion module filters the lidar point cloud data to remove invalid points, performs Gaussian filtering or median filtering on the camera image, calibrates the lidar's XYZ coordinate system with the camera's image coordinate system, unifies the timestamps, aligns asynchronously acquired data through interpolation, converts the lidar point cloud into a BEV (Browser Elevation Vehicle), converts the camera image into a grayscale image, projects the lidar point cloud onto the camera image plane, marks obstacle positions, extracts the geometric features of the lidar point cloud and the semantic features of the camera image, and fuses them using Kalman filtering.

[0015] Preferably, the path planning module generates an obstacle avoidance path based on the obstacle location and dynamic trajectory data provided by the multi-source data acquisition and fusion module, combined with the target allocated by the task scheduling module, through a hybrid path planning algorithm. It integrates the obstacle location and trajectory data from the LiDAR and visual camera sensors into a dynamic occupancy grid map, marks static and dynamic obstacles, predicts the trajectory of dynamic obstacles, and generates possible locations for the next few frames.

[0016] Preferably, the path planning module divides the warehouse map into a global grid map and a local grid map. In the local map, the obstacle positions are updated based on real-time sensor data. Based on the target assigned by the task scheduling module, an initial collision-free path is generated, as shown in the formula:

[0017]

[0018] In the formula, C dc (t) represents the dynamic environmental cost at the current time step t, N represents the number of dynamic obstacles in the environment, and d i (t) represents the Euclidean distance between the robot and the i-th dynamic obstacle at time t, where r i (t) represents the predicted trajectory radius of the i-th dynamic obstacle at time t, τ i (t) represents the motion time constant of the i-th dynamic obstacle at time t, where T se This represents the preset safety time threshold, and α and β represent weighting coefficients that balance the priority between distance cost and time cost.

[0019] The real-time impact of obstacles in the quantified environment on path planning is considered, taking into account both distance and time risks. The closer the obstacle or the more violent its movement, the higher the cost.

[0020] Preferably, the path planning module calculates feasible control commands based on the robot's current speed and acceleration, selects the command that minimizes the path cost, decomposes the path into a series of time-interval posture sequences, adjusts the trajectory through an optimization algorithm to avoid obstacles, and the overall optimization objective function of the path is expressed by the formula:

[0021]

[0022] In the formula, E oe The overall optimization objective function represents the path. This indicates that for the time interval t0, t f Integrate along the path within the given path, k(t). 2 Let μ represent the curvature of the path at time t, and let μ represent the weighting coefficient of the rate of change of velocity. Let v(t) represent the rate of change of velocity, v(t) represent the robot's velocity at time t, and γ represent the balance factor, where t0, t... f Represents the start and end points of the time interval for path planning, (1-γ)·C dc (t) represents the path optimization introduced by the dynamic environment cost, ensuring that the path avoids dynamic obstacles, optimizing the smoothness of the path and energy efficiency, by minimizing the square of curvature and the square of the rate of change of velocity, sharing the path information of each robot through the multi-robot collaboration module, detecting potential path intersections or time overlaps, using a spatiotemporal graph to represent the robot path, identifying possible collision points, and dynamically allocating robot priorities according to the urgency of the task, with high-priority robots having priority passage.

[0023] Preferably, in the multi-robot collaborative module, each robot collects its own status information in real time through sensors, including current position, speed, direction, target point, path planning results, and dynamic constraints. It broadcasts or sends its own status information point-to-point to other robots and the central coordination node via a distributed communication protocol. Based on the current path planning results and its own status, it predicts its motion trajectory over a future period, divides the workspace into grids or spatial partitions for quickly locating potential conflict areas, and prioritizes conflicts by comparing its own trajectory with those of other robots based on conflict type, distance, and time urgency. It assigns priorities to robots based on task importance, path length, and energy consumption. After adjusting the strategy, it updates its own status information in real time and broadcasts it to other robots via the communication protocol.

[0024] Preferably, the execution control module obtains the target position, target direction, and path trajectory from the path planning module, decomposes the path command into target values ​​of linear velocity and angular velocity, acquires the robot's current position, velocity, and direction in real time through sensors, compares them with the target values, calculates the deviation, outputs the PID controller, calculates the target velocity of the left and right wheels based on the linear velocity and angular velocity output by the PID controller through a differential drive algorithm, sends the left and right wheel speed commands to the drive motors, drives the robot to move along the planned path, compares the real-time feedback with the target path, recalculates the error and adjusts the PID parameters, and if a severe deviation from the path is detected, triggers emergency braking or switches to obstacle avoidance mode.

[0025] Preferably, the task scheduling module receives task instructions from the warehouse management system, parses key information such as task type, target location, priority, and time constraints, obtains path information generated by the path planning module for the task, evaluates the executability of the task, collects the status of all robots in real time, checks whether the task path conflicts with other robot paths, marks high-risk tasks, sorts the task queue according to task priority and path efficiency, matches the optimal robot for each task, binds the task to the target robot, generates a unique task identifier, records the task start time and estimated completion time, sends task instructions to the robot, and synchronously updates the task queue and robot status. If robot execution is interrupted, task reassignment and path replanning are triggered.

[0026] Preferably, the safety monitoring module receives sensor data from the multi-source data acquisition module and robot status feedback from the execution control module. Based on historical data, it establishes a normal behavior model of hardware operating status and environmental parameters. It compares the current sensor data with the model to detect abnormal patterns. According to the type and scope of the abnormality, it triggers different levels of early warning signals to determine whether it is a recoverable fault or a permanent fault requiring manual intervention. It analyzes the impact of external environmental anomalies on robot safety, dynamically adjusts the risk level in conjunction with the path planning module, and transmits the detected anomaly type, occurrence time, and related parameters to the self-optimization module. The self-optimization module calls a strategy iteration algorithm to recalculate the control parameters based on the current anomaly information.

[0027] Compared with the prior art, the beneficial effects of the present invention are:

[0028] 1. This invention integrates LiDAR and visual data into a dynamic occupancy grid map through a hybrid path planning algorithm and introduces a dynamic obstacle trajectory prediction model to achieve active avoidance of moving obstacles. By quantifying both distance and time costs, path planning can optimize path smoothness and energy efficiency while ensuring safe distance. Furthermore, the path is decomposed into time-interval posture sequences and combined with curvature optimization objective function, improving the robot's ability to maintain efficient and stable obstacle avoidance capabilities even in densely populated or frequently moving warehouse environments, significantly increasing task success rate.

[0029] 2. This invention uses a distributed communication protocol to quickly locate potential conflict areas through spatial partitioning and trajectory prediction, and dynamically allocates right-of-way based on task priority. Spatiotemporal graph modeling can detect path intersections in real time. Combined with the priority ranking strategy for high-priority tasks, it avoids low-priority robots from waiting for a long time. The priority ranking mechanism for trajectory conflicts can dynamically adjust the avoidance strategy, reduce the number of path replanning, and significantly improve the overall task throughput and system robustness. Attached Figure Description

[0030] Figure 1 This is a schematic diagram of the structure of an intelligent obstacle avoidance system for warehouse robots under the Internet of Things according to the present invention;

[0031] Figure 2 This invention provides an operational flow diagram of an intelligent obstacle avoidance system for warehouse robots under the Internet of Things (IoT). Figure 1 ;

[0032] Figure 3 This invention provides an operational flow diagram of an intelligent obstacle avoidance system for warehouse robots under the Internet of Things (IoT). Figure 2 . Detailed Implementation

[0033] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0034] Example

[0035] Please see Figures 1-3 As shown, the present invention provides a technical solution including a multi-source data acquisition and fusion module, a path planning module, a multi-robot collaboration module, an execution control module, a task scheduling module, a safety monitoring module, and a self-optimization module;

[0036] The multi-source data acquisition and fusion module integrates environmental obstacle information based on the raw data from lidar and visual camera sensors through a multi-sensor adaptive fusion algorithm.

[0037] The path planning module generates obstacle avoidance paths based on obstacle location and dynamic trajectory data provided by the multi-source data acquisition and fusion module, combined with the target assigned by the task scheduling module, using a hybrid path planning algorithm.

[0038] The multi-robot collaboration module performs path conflict detection and collaborative adjustment based on the path information from the path planning module and the real-time status shared with other robots through a distributed communication protocol.

[0039] The execution control module controls the robot's motion posture through a PID controller and a differential drive system, based on the obstacle avoidance path instructions output by the path planning module.

[0040] The task scheduling module allocates robot task targets based on the task requirements issued by the warehouse management system and the path information generated by the path planning module through a dynamic allocation algorithm.

[0041] The security monitoring module, based on the sensor data from the multi-source data acquisition and fusion module and the feedback status from the execution control module, uses an anomaly detection mechanism to provide early warnings of hardware failures and environmental risks, and triggers strategy iterations of the self-optimization module when an anomaly occurs.

[0042] The self-optimization module iteratively updates the obstacle avoidance strategy and path planning algorithm based on the abnormal feedback from the safety monitoring module and the historical obstacle avoidance data from the path planning module through a reinforcement learning model.

[0043] Preferably, the multi-source data acquisition and fusion module integrates environmental obstacle information based on the raw data from the lidar and visual camera sensors through a multi-sensor adaptive fusion algorithm; by selecting lidar and visual camera as core data sources, configuring the communication protocols and data formats of each sensor, and setting the sampling frequency, resolution, and timestamp synchronization mechanism of the sensors.

[0044] Preferably, the multi-source data acquisition and fusion module filters the lidar point cloud data to remove invalid points, performs Gaussian filtering or median filtering on the camera image, calibrates the lidar's XYZ coordinate system with the camera's image coordinate system, unifies the timestamps, aligns asynchronously acquired data through interpolation, converts the lidar point cloud into a BEV (Browser Elevation Vehicle), converts the camera image into a grayscale image, projects the lidar point cloud onto the camera image plane, marks obstacle positions, extracts the geometric features of the lidar point cloud and the semantic features of the camera image, and fuses them using Kalman filtering.

[0045] Preferably, the path planning module generates an obstacle avoidance path based on the obstacle location and dynamic trajectory data provided by the multi-source data acquisition and fusion module, combined with the target allocated by the task scheduling module, through a hybrid path planning algorithm. It integrates the obstacle location and trajectory data from the LiDAR and visual camera sensors into a dynamic occupancy grid map, marks static and dynamic obstacles, predicts the trajectory of dynamic obstacles, and generates possible locations for the next few frames.

[0046] Preferably, the path planning module divides the warehouse map into a global grid map and a local grid map. In the local map, the obstacle positions are updated based on real-time sensor data. Based on the target assigned by the task scheduling module, an initial collision-free path is generated, as shown in the formula:

[0047]

[0048] In the formula, C dc (t) represents the dynamic environmental cost at the current time step t, N represents the number of dynamic obstacles in the environment, and d i (t) represents the Euclidean distance between the robot and the i-th dynamic obstacle at time t, where r i (t) represents the predicted trajectory radius of the i-th dynamic obstacle at time t, τ i (t) represents the motion time constant of the i-th dynamic obstacle at time t, where T se This represents the preset safety time threshold, and α and β represent weighting coefficients that balance the priority between distance cost and time cost.

[0049] The real-time impact of obstacles in the quantified environment on path planning is considered, taking into account both distance and time risks. The closer the obstacle or the more violent its movement, the higher the cost.

[0050] Preferably, the path planning module calculates feasible control commands based on the robot's current speed and acceleration, selects the command that minimizes the path cost, decomposes the path into a series of time-interval posture sequences, adjusts the trajectory through an optimization algorithm to avoid obstacles, and the overall optimization objective function of the path is expressed by the formula:

[0051]

[0052] In the formula, E oe The overall optimization objective function represents the path. This indicates that for the time interval t0, t f Integrate along the path within the given path, k(t). 2 Let μ represent the curvature of the path at time t, and let μ represent the weighting coefficient of the rate of change of velocity. Let v(t) represent the rate of change of velocity, v(t) represent the robot's velocity at time t, and γ represent the balance factor, where t0, t... f Represents the start and end points of the time interval for path planning, (1-γ)·C dc (t) represents the path optimization introduced by the dynamic environment cost, ensuring that the path avoids dynamic obstacles, optimizing the smoothness of the path and energy efficiency, by minimizing the square of curvature and the square of the rate of change of velocity, sharing the path information of each robot through the multi-robot collaboration module, detecting potential path intersections or time overlaps, using a spatiotemporal graph to represent the robot path, identifying possible collision points, and dynamically allocating robot priorities according to the urgency of the task, with high-priority robots having priority passage.

[0053] Preferably, in the multi-robot collaborative module, each robot collects its own status information in real time through sensors, including current position, speed, direction, target point, path planning results, and dynamic constraints. It broadcasts or sends its own status information point-to-point to other robots and the central coordination node via a distributed communication protocol. Based on the current path planning results and its own status, it predicts its motion trajectory over a future period, divides the workspace into grids or spatial partitions for quickly locating potential conflict areas, and prioritizes conflicts by comparing its own trajectory with those of other robots based on conflict type, distance, and time urgency. It assigns priorities to robots based on task importance, path length, and energy consumption. After adjusting the strategy, it updates its own status information in real time and broadcasts it to other robots via the communication protocol.

[0054] Preferably, the execution control module obtains the target position, target direction, and path trajectory from the path planning module, decomposes the path command into target values ​​of linear velocity and angular velocity, acquires the robot's current position, velocity, and direction in real time through sensors, compares them with the target values, calculates the deviation, outputs the PID controller, calculates the target velocity of the left and right wheels based on the linear velocity and angular velocity output by the PID controller through a differential drive algorithm, sends the left and right wheel speed commands to the drive motors, drives the robot to move along the planned path, compares the real-time feedback with the target path, recalculates the error and adjusts the PID parameters, and if a severe deviation from the path is detected, triggers emergency braking or switches to obstacle avoidance mode.

[0055] Preferably, the task scheduling module receives task instructions from the warehouse management system, parses key information such as task type, target location, priority, and time constraints, obtains path information generated by the path planning module for the task, evaluates the executability of the task, collects the status of all robots in real time, checks whether the task path conflicts with other robot paths, marks high-risk tasks, sorts the task queue according to task priority and path efficiency, matches the optimal robot for each task, binds the task to the target robot, generates a unique task identifier, records the task start time and estimated completion time, sends task instructions to the robot, and synchronously updates the task queue and robot status. If robot execution is interrupted, task reassignment and path replanning are triggered.

[0056] Preferably, the safety monitoring module receives sensor data from the multi-source data acquisition module and robot status feedback from the execution control module. Based on historical data, it establishes a normal behavior model of hardware operating status and environmental parameters. It compares the current sensor data with the model to detect abnormal patterns. According to the type and scope of the abnormality, it triggers different levels of early warning signals to determine whether it is a recoverable fault or a permanent fault requiring manual intervention. It analyzes the impact of external environmental anomalies on robot safety, dynamically adjusts the risk level in conjunction with the path planning module, and transmits the detected anomaly type, occurrence time, and related parameters to the self-optimization module. The self-optimization module calls a strategy iteration algorithm to recalculate the control parameters based on the current anomaly information.

[0057] Working Principle: Environmental data is collected in real time through sensors such as LiDAR and visual cameras. A multi-sensor adaptive fusion algorithm is used to standardize, denoise, and spatiotemporally align heterogeneous data. LiDAR point cloud data and camera images are integrated into unified obstacle location information through coordinate system calibration and projection transformation. Geometric and semantic features are extracted. Based on obstacle location and dynamic trajectory data provided by the multi-source data acquisition module, and combined with targets assigned by the task scheduling module, a dynamic occupancy grid map is constructed. A hybrid path planning algorithm is used to divide the map into a global grid map and a local grid map, and the obstacle locations and dynamic obstacle trajectories in the local map are updated in real time. The model generates possible future locations, quantifies distance and time risk costs, and comprehensively generates collision-free and efficient obstacle avoidance paths. It also optimizes trajectory smoothness and energy efficiency through algorithms, sharing real-time states of each robot via a distributed communication protocol. Based on spatial partitioning and trajectory prediction, it quickly locates potential conflict areas and identifies path intersections or temporal overlaps through spatiotemporal graph modeling. The avoidance strategy is dynamically adjusted based on factors such as conflict type, distance, and task priority, with high-priority robots having priority passage. After adjustments, the robot's status information is updated in real-time and broadcast. The system receives obstacle avoidance path instructions from the path planning module, decomposes them into linear and angular velocity target values, and calculates the deviation using a PID controller. The system outputs control signals and, combined with a differential drive algorithm, generates speed commands for the left and right wheels. The drive motor moves along the planned path, and real-time feedback sensor data is compared with the target path. PID parameters are dynamically adjusted to optimize tracking accuracy. If a significant deviation from the path is detected, emergency braking is immediately triggered or the system switches to obstacle avoidance mode. The task scheduling module receives task commands from the warehouse management system and, combined with path information from the path planning module, assesses task feasibility. Through a dynamic allocation algorithm, the optimal robot is matched to the task, and a unique task identifier is assigned. The system monitors robot status and path conflicts in real time, marking high-risk tasks and triggering reassignment. If a task is interrupted, path replanning and task reassignment are automatically triggered, ensuring safety monitoring. The module integrates multi-source sensor data and feedback status from the execution control module. Based on historical data, it establishes a normal behavior model of hardware and environment, detects abnormal patterns through real-time comparison, triggers graded early warnings based on the type and scope of the abnormality, analyzes hardware failures and environmental risks, dynamically adjusts the risk level of path planning, and transmits abnormal information to the self-optimization module. Based on the abnormal feedback from the safety monitoring module and historical obstacle avoidance data of path planning, it iteratively updates the obstacle avoidance strategy and path planning algorithm through a reinforcement learning model. Reinforcement learning learns the optimal decision strategy autonomously through a reward and punishment mechanism, and dynamically adjusts the dynamic environmental cost weight. The optimized control parameters and path algorithm are then sent to the execution module in real time.

[0058] Although embodiments of the invention have been shown and described, it will be understood by those skilled in the art that various changes, modifications, substitutions and alterations can be made to these embodiments without departing from the principles and spirit of the invention, the scope of which is defined by the appended claims and their likenesses.

[0059] The present invention and its embodiments have been described above. This description is not restrictive, and the accompanying drawings are only one embodiment of the present invention; the actual structure is not limited thereto. In conclusion, if those skilled in the art are inspired by this description and design similar structures and embodiments without departing from the spirit of the invention, such designs should fall within the protection scope of the present invention.

Claims

1. An intelligent obstacle avoidance system for warehouse robots under the Internet of Things, characterized in that: It includes a multi-source data acquisition and fusion module, a path planning module, a multi-robot collaboration module, an execution control module, a task scheduling module, a safety monitoring module, and a self-optimization module; The multi-source data acquisition and fusion module integrates environmental obstacle information based on the raw data from lidar and visual camera sensors through a multi-sensor adaptive fusion algorithm. The path planning module generates obstacle avoidance paths based on obstacle location and dynamic trajectory data provided by the multi-source data acquisition and fusion module, combined with the target assigned by the task scheduling module, using a hybrid path planning algorithm. The multi-robot collaboration module performs path conflict detection and collaborative adjustment based on the path information from the path planning module and the real-time status shared with other robots through a distributed communication protocol. The execution control module controls the robot's motion posture through a PID controller and a differential drive system, based on the obstacle avoidance path instructions output by the path planning module. The task scheduling module allocates robot task targets based on the task requirements issued by the warehouse management system and the path information generated by the path planning module through a dynamic allocation algorithm. The security monitoring module, based on the sensor data from the multi-source data acquisition and fusion module and the feedback status from the execution control module, uses an anomaly detection mechanism to provide early warnings of hardware failures and environmental risks, and triggers strategy iterations of the self-optimization module when an anomaly occurs. The self-optimization module iteratively updates the obstacle avoidance strategy and path planning algorithm based on the abnormal feedback from the safety monitoring module and the historical obstacle avoidance data from the path planning module through a reinforcement learning model.

2. The intelligent obstacle avoidance system for warehouse robots under the Internet of Things as described in claim 1, characterized in that: The multi-source data acquisition and fusion module integrates environmental obstacle information based on the raw data from lidar and visual camera sensors through a multi-sensor adaptive fusion algorithm. By selecting LiDAR and vision cameras as core data sources, configuring the communication protocols and data formats of each sensor, and setting the sampling frequency, resolution, and timestamp synchronization mechanism of the sensors.

3. The intelligent obstacle avoidance system for warehouse robots under the Internet of Things according to claim 2, characterized in that: The multi-source data acquisition and fusion module filters the LiDAR point cloud data to remove invalid points, applies Gaussian or median filtering to the camera images, calibrates the LiDAR's XYZ coordinate system with the camera's image coordinate system, unifies timestamps, aligns asynchronously acquired data through interpolation, converts the LiDAR point cloud into a BEV (Browser Elevation Vehicle), converts the camera images into grayscale images, projects the LiDAR point cloud onto the camera image plane, marks obstacle positions, extracts the geometric features of the LiDAR point cloud and the semantic features of the camera images, and fuses them using Kalman filtering.

4. The intelligent obstacle avoidance system for warehouse robots under the Internet of Things as described in claim 1, characterized in that: The path planning module generates obstacle avoidance paths by combining obstacle location and dynamic trajectory data provided by the multi-source data acquisition and fusion module with the target assigned by the task scheduling module through a hybrid path planning algorithm. It integrates obstacle location and trajectory data from LiDAR and visual camera sensors into a dynamic occupancy grid map, marks static and dynamic obstacles, predicts the trajectory of dynamic obstacles, and generates possible locations for the next few frames.

5. The intelligent obstacle avoidance system for warehouse robots under the Internet of Things according to claim 4, characterized in that: The path planning module divides the warehouse map into a global grid map and a local grid map. In the local map, it updates obstacle positions based on real-time sensor data and generates an initial collision-free path, as shown in the formula: In the formula, C dc (t) represents the dynamic environmental cost at the current time step t, N represents the number of dynamic obstacles in the environment, and d i (t) represents the Euclidean distance between the robot and the i-th dynamic obstacle at time t, where r i (t) represents the predicted trajectory radius of the i-th dynamic obstacle at time t, τ i (t) represents the motion time constant of the i-th dynamic obstacle at time t, where T se This represents the preset safety time threshold, and α and β represent weighting coefficients that balance the priority between distance cost and time cost.

6. The intelligent obstacle avoidance system for warehouse robots under the Internet of Things as described in claim 5, characterized in that: The path planning module calculates feasible control commands based on the robot's current speed and acceleration, selects the command that minimizes the path cost, decomposes the path into a series of time-interval posture sequences, adjusts the trajectory through an optimization algorithm to avoid obstacles, and the overall optimization objective function of the path is expressed by the formula: In the formula, E oe The overall optimization objective function represents the path. This indicates that for the time interval t0, t f Integrate along the path within the given path, k(t). 2 Let μ represent the curvature of the path at time t, and let μ represent the weighting coefficient of the rate of change of velocity. Let v(t) represent the rate of change of velocity, v(t) represent the robot's velocity at time t, and γ represent the balance factor, where t0, t... f Represents the start and end points of the time interval for path planning, (1-γ)·C dc (t) represents the path optimization introduced by the dynamic environment cost, ensuring that the path avoids dynamic obstacles.

7. The intelligent obstacle avoidance system for warehouse robots under the Internet of Things according to claim 1, characterized in that: The multi-robot collaborative module allows each robot to collect its own status information in real time through sensors, including current position, speed, direction, target point, path planning results, and dynamic constraints. It broadcasts or sends its status information point-to-point to other robots and the central coordination node via a distributed communication protocol. Based on the current path planning results and its own status, it predicts its trajectory over a future period, divides the workspace into grids or spatial partitions for rapid location of potential conflict areas, and prioritizes conflicts by comparing its own trajectory with those of other robots, based on conflict type, distance, and time urgency. It also assigns priorities to robots based on task importance, path length, and energy consumption. After adjusting its strategy, it updates its own status information in real time and broadcasts it to other robots via the communication protocol.

8. The intelligent obstacle avoidance system for warehouse robots under the Internet of Things according to claim 1, characterized in that: The execution control module obtains the target position, target direction, and path trajectory from the path planning module, decomposes the path command into target values ​​of linear velocity and angular velocity, acquires the robot's current position, velocity, and direction in real time through sensors, compares them with the target values, calculates the deviation, and outputs the PID controller. Based on the linear velocity and angular velocity output by the PID controller, the target velocity of the left and right wheels is calculated through the differential drive algorithm, and the speed commands of the left and right wheels are sent to the drive motors to drive the robot to move along the planned path. The real-time feedback is compared with the target path, the error is recalculated, and the PID parameters are adjusted. If a serious deviation from the path is detected, emergency braking is triggered or the robot switches to obstacle avoidance mode.

9. The intelligent obstacle avoidance system for warehouse robots under the Internet of Things according to claim 1, characterized in that: The task scheduling module receives task instructions from the warehouse management system, parses key information such as task type, target location, priority, and time constraints, obtains path information generated by the path planning module for the task, evaluates the executability of the task, collects the status of all robots in real time, checks whether the task path conflicts with other robot paths, marks high-risk tasks, sorts the task queue according to task priority and path efficiency, matches the optimal robot for each task, binds the task to the target robot, generates a unique task identifier, records the task start time and estimated completion time, sends task instructions to the robot, and synchronously updates the task queue and robot status. If robot execution is interrupted, task reassignment and path replanning are triggered.

10. The intelligent obstacle avoidance system for warehouse robots under the Internet of Things according to claim 1, characterized in that: The safety monitoring module receives sensor data from the multi-source data acquisition module and robot status feedback from the execution control module. Based on historical data, it establishes a normal behavior model of hardware operating status and environmental parameters. It compares the current sensor data with the model to detect abnormal patterns. According to the type and scope of the abnormality, it triggers different levels of early warning signals to determine whether it is a recoverable fault or a permanent fault requiring manual intervention. It analyzes the impact of external environmental anomalies on robot safety and dynamically adjusts the risk level in conjunction with the path planning module. It transmits the detected anomaly type, occurrence time, and related parameters to the self-optimization module. Based on the current anomaly information, the self-optimization module calls the strategy iteration algorithm to recalculate the control parameters.

Images