Warehouse inventory robot dynamic window planning method fusing rfid signal perception

Abstract

This invention discloses a dynamic window planning method for warehouse inventory robots that integrates RFID signal sensing. First, a physical dynamic window is constructed, which is the intersection of a set of speed limits, a set of dynamic acceleration limits, and a set of safety braking limits. Within this physical dynamic window, multiple candidate speeds are generated through discrete sampling. A five-dimensional comprehensive evaluation function is used to score each candidate speed. This function includes azimuth angle, safety distance, speed, RFID signal gain prediction, and energy consumption penalty. Finally, the speed command with the highest comprehensive score is selected and sent to the robot chassis for execution. Furthermore, this invention introduces a weight adaptive adjustment mechanism, which dynamically adjusts the weights of various indicators in the comprehensive evaluation function based on the robot's remaining battery power and the value of the inventory area. This approach ensures robot obstacle avoidance safety while guiding the robot to avoid RFID signal blind spots and reducing operating energy consumption, thus balancing multiple needs.

Description

Technical Field

[0001] This invention belongs to the field of mobile robot path planning technology, specifically involving a dynamic window planning method for warehouse inventory robots that integrates RFID signal perception, which is applicable to complex indoor warehouse inventory scenarios where there is signal shielding from metal shelves. Background Technology

[0002] In current smart warehousing scenarios for e-commerce and third-party logistics, warehouse inventory robots equipped with RFID readers have become core equipment for in-warehouse inventory operations. These robots autonomously inspect and read shelf labels and verify inventory data, significantly reducing the labor costs of manual inventory checks while improving the accuracy of inventory data, thus driving the transformation of warehouse management towards automation and intelligence.

[0003] However, in practical warehousing applications, existing inventory robots generally employ a hierarchical path planning scheme combining the A* algorithm and the Dynamic Window Method (DWA). These schemes largely rely on static environment models for design and fail to adequately adapt to actual on-site conditions. Temporarily stacked goods and moving personnel in warehouses create dynamic obstacles; metal shelves shield RFID signals and cause multipath reflections; and robot endurance and energy consumption constraints directly impact operational continuity. These factors are not considered in existing planning methods.

[0004] Faced with the aforementioned practical problems, existing path planning schemes struggle to make rapid and reasonable adjustments. A* replanning exhibits slow response times, and zigzag paths are prone to causing robot motion jitter. DWA obstacle avoidance fails to account for RFID signal coverage, potentially leading the robot into signal blind spots and causing tag misreads. Furthermore, unconstrained sharp turns exacerbate energy consumption. This not only increases inventory time but also affects robot operational stability and data reading reliability.

[0005] Based on this, the present invention optimizes and improves the traditional dynamic window method to meet the core operational requirements of warehouse inventory robots, and incorporates RFID signal perception, energy consumption constraints and obstacle avoidance safety coordination into the path planning decision system, thereby solving the above-mentioned defects of the existing technology. Summary of the Invention

[0006] To address the aforementioned issues, this invention discloses a dynamic window planning method for warehouse inventory robots that integrates RFID signal sensing. This method improves the inventory read rate, reduces robot energy consumption, and extends the lifespan of the motor while ensuring obstacle avoidance safety.

[0007] To achieve the above objectives, the technical solution of the present invention is as follows:

[0008] A dynamic window planning method for warehouse inventory robots integrating RFID signal sensing includes the following steps:

[0009] S1: First, establish the differential drive kinematic model of the robot to understand the robot's response to a given linear velocity ( ) and angular velocity ( How does the pose change recursively under these conditions?

[0010] S2: Combining robot physical limits and safety constraints, construct a dynamic velocity sampling window formed by the intersection of a set of velocity limits, a set of dynamic acceleration limits, and a set of safety braking limits. );

[0011] S3: In this dynamic window ( The linear velocity () ) and angular velocity ( Discrete sampling is performed to generate multiple sets of candidate velocity commands. The trajectory of each candidate velocity is simulated and predicted. Then, a five-dimensional comprehensive evaluation function is used, which includes azimuth evaluation, safety distance evaluation, velocity evaluation, RFID signal gain prediction, and energy consumption penalty. Each predicted trajectory is scored; the weights of each component of the evaluation function can be dynamically and adaptively adjusted according to the working conditions.

[0012] S4: Finally, select the comprehensive evaluation function. The candidate speed command with the maximum value is obtained and converted into a motor control signal, which is then sent to the robot chassis for execution.

[0013] The beneficial effects of this invention are as follows:

[0014] (1) Real-time performance and smoothness: Based on the speed sampling of the physical dynamic window, the single planning cycle can be compressed to the millisecond level, and the output speed strictly conforms to the acceleration and deceleration limits of the motor. The trajectory is naturally smooth, which can effectively avoid the situation of motor step jitter, thereby extending the life of the transmission mechanism.

[0015] (2) Obstacle avoidance and business collaboration: By adding an RFID signal gain evaluation item, the robot will actively predict and drive towards the high confidence area of ​​the signal when it is going around dynamic obstacles, which can effectively avoid the signal blind spot caused by metal shelves and improve the success rate of inventory reading.

[0016] (3) Energy consumption optimization: Energy consumption penalty can suppress the behavior of sharp turns in place, and reduce the energy consumption caused by detours while ensuring safety.

[0017] (4) Scene Adaptation: The weight adjustment mechanism enables the algorithm to automatically adjust its behavior strategy based on the robot's power status and the value of the inventory task, thereby improving the robustness of the system. Attached Figure Description

[0018] Figure 1This is an overall flowchart of the present invention;

[0019] Figure 2 This is a schematic diagram illustrating the application of RFID-guided obstacle avoidance in a warehouse passageway scenario according to the present invention. Detailed Implementation

[0020] The present invention will be further illustrated below with reference to the accompanying drawings and specific embodiments. It should be understood that the following specific embodiments are for illustrative purposes only and are not intended to limit the scope of the invention.

[0021] The present invention discloses a dynamic window planning method for a warehouse inventory robot that integrates RFID signal sensing. The overall implementation process is as follows: Figure 1 As shown, a kinematic model is first established for the differential drive robot, and then a dynamic window is constructed in combination with motor hardware constraints to limit the feasible speed range within each control cycle. Then, candidate speeds are generated by sampling within the window, and the predicted trajectory is scored by an evaluation function that integrates multiple dimensions such as RFID signals and energy consumption. Finally, the speed command with the highest score is selected and sent to the chassis for execution. Figure 2 This is a schematic diagram of RFID-guided obstacle avoidance in a warehouse passage scenario according to the present invention. The above process will be described in detail below with reference to specific embodiments.

[0022] I. Construction of Dynamic Windows

[0023] In practice, by combining the motion characteristics of the differential-driven warehouse inventory robot with the physical constraints of the motor, the executable speed sampling range within a single control cycle is determined, providing a safe and feasible constraint boundary for subsequent trajectory sampling.

[0024] 1. Robot kinematic modeling

[0025] For a differential-driven robot moving in a two-dimensional plane, its pose is determined by coordinates in the world coordinate system. With heading angle Description. At a given linear velocity and angular velocity Under the control input, after a time step New position after We obtain it from the kinematic integral:

[0026] ;

[0027] d is the integral operator, and τ is an integral variable used to represent any time point from t to t+Δt.

[0028] In actual control processes, the control cycle is taken into consideration. The period is relatively short, so we can assume the linear velocity within the period is relatively short. and angular velocity To keep it constant, the continuous model is simplified to a discrete form. ;

[0029] The physical meanings of each parameter are as follows: Linear velocity The forward / backward speed of the robot chassis along the current heading, measured in m / s, with positive values ​​for forward and negative values ​​for backward; angular velocity. The rotational speed of the robot about its own vertical axis, measured in rad / s; positive values ​​indicate counter-clockwise rotation, and negative values ​​indicate clockwise rotation; yaw angle. The angle between the robot's front orientation and the positive x-axis of the world coordinate system, measured in rad; time step. The control period between two state updates is measured in seconds (s).

[0030] 2. Construction of the speed limit set (Vs)

[0031] Based on the robot's hardware parameters, the maximum linear velocity in this embodiment is set to 1.5 m / s. The physical limits of the chassis motors directly determine the robot's executable speed range. The feasible range for linear velocity is [-1.5, 1.5] m / s, and the feasible range for angular velocity is [-0.8, 0.8] rad / s. The speed limit set ( The mathematical expression for ) is: This set strictly constrains the robot's speed at any time, ensuring it does not exceed the hardware output limit of the motor, and forms the fundamental physical boundary of the dynamic window algorithm.

[0032] 3. Dynamic acceleration limit set ( ) build

[0033] The robot's acceleration and deceleration capabilities are determined by the physical properties of its motors. In this embodiment, the hardware parameters are: maximum linear acceleration... m / s², maximum linear deceleration m / s², maximum angular acceleration rad / s², maximum angular deceleration rad / s². Control period. The time is usually 0.1s.

[0034] exist Within a given time period, the robot's velocity change is limited by the aforementioned acceleration and deceleration capabilities. Let the current linear velocity be... The current angular velocity is Then the feasible region of velocity at the next moment is:

[0035] ;

[0036] This set demonstrates the robot's ability to change speed in a short period of time, avoiding control commands that exceed the motor's load.

[0037] 4. Safety braking limit set ( ) build

[0038] To ensure the robot does not collide with obstacles during operation, the braking distance must not exceed the safe distance from the obstacle. According to the formula for uniformly decelerated motion, the braking distance... ,in This is the maximum braking deceleration (3.5 m / s² in this embodiment). This refers to the distance to the nearest obstacle as measured by the lidar. The safety constraints are: The maximum safe linear velocity is obtained by sorting: Substituting the braking deceleration parameters of this embodiment, it can be simplified to: ; Safety braking limit set ( The mathematical expression for ) is: The system constrains the robot's speed from a safety perspective, ensuring that it can stop in front of obstacles under any circumstances.

[0039] 5. Dynamic windows ( The formation of

[0040] Considering the above constraints, the velocity sampling space (i.e., the dynamic window) available to the robot in the next moment is the intersection of three sets: This dynamic window simultaneously meets the hardware limits of the motor, the physical constraints of acceleration and deceleration, and the requirements for collision avoidance safety. It strictly follows Newton's laws of motion and the physical limits of the motor, providing an absolutely safe and executable two-dimensional velocity search space for subsequent trajectory sampling.

[0041] II. Construction of Multidimensional Evaluation Function and Trajectory Scoring

[0042] After determining the feasible velocity sampling range, the optimal command is further selected from the feasible velocity space through an evaluation function. Candidate velocities are generated by sampling within a dynamic window, and a five-dimensional comprehensive evaluation function is used to integrate azimuth angle, safety distance, velocity, RFID signal gain, and energy consumption penalty. The predicted trajectories are scored, and trajectories with good signal quality and low energy consumption are given priority.

[0043] 1. Discretized velocity sampling and trajectory prediction

[0044] In practice, the linear velocity step size is set according to the resolution of the robot chassis controller. and angular velocity step In dynamic windows ( ) Perform gridded sampling within the range to generate N candidate velocity pairs .

[0045] For each group of candidate speeds Assuming the robot is in a future prediction time window Maintaining a constant speed, the trajectory's end pose is predicted based on a discrete-time kinematic model. : In the formula, This represents the robot's current absolute pose.

[0046] 2. Construction of Multidimensional Comprehensive Evaluation Function

[0047] Based on the actual business needs of warehouse inventory counting, this invention proposes an improved DWA comprehensive evaluation function for warehouse inventory counting scenarios. Its calculation expression is as follows:

[0048] ;

[0049] in, , , , , These are the weighting coefficients for each item. This is a smoothing normalization constant. The specific calculation logic for each evaluation item is as follows:

[0050] Azimuth evaluation item This item corresponds to the angle between the predicted trajectory's end and the line connecting it to the global target point. The smaller the angle, the higher the score, which is used to guide the robot to always maintain a convergent motion trend toward the target point.

[0051] Safe distance evaluation items This corresponds to the distance between the predicted trajectory and the nearest obstacle in the environment, including both dynamic workers and static shelves. If this distance is less than the set physical safety radius... Then directly calculate the total score of the speed sample. A score of 0 is forcibly assigned, and a veto mechanism is used to ensure absolute collision avoidance; if the distance is greater than the safe radius, the greater the distance, the higher the score.

[0052] Speed ​​evaluation item To ensure the overall efficiency of the inventory operation, this item is directly related to the current sampling linear velocity. It is directly proportional, and the specific calculation method is as follows: .

[0053] RFID signal gain prediction item The core function of the inventory robot is reading RFID tags. This invention pre-establishes a two-dimensional RFID signal confidence grid map based on multipath effect calculations. When the predicted coordinates are derived... Then, query the map to obtain the corresponding expected signal strength. And normalize it: In the formula and These represent the maximum and minimum signal strength values ​​from historical statistics. Using this information, the scoring system will prioritize trajectories leading to areas with high signal confidence, allowing the robot to proactively avoid signal blind spots caused by metal shelves during obstacle avoidance.

[0054] Energy consumption penalty evaluation items Differential drive robots generate significant motor energy consumption when making high angular velocity turns. The specific calculation method for this is as follows: The formula shows that the smaller the absolute value of the angular velocity, the higher the score for this item, which is used to guide the robot to use smooth curves to reduce energy-intensive sharp turns in place.

[0055] 3. Dynamic adaptive adjustment of weights

[0056] This invention establishes a weight adaptive mechanism based on environmental conditions, which can dynamically adjust the weight allocation according to the robot's working conditions and tasks:

[0057] When the system detects that the robot's remaining battery power is below a preset threshold (e.g., 20%), it will automatically increase the energy consumption weight. Strengthen energy-saving detour strategies;

[0058] When the robot enters the key high-value goods inventory area and needs to ensure an extremely high read rate, the system automatically increases the signal gain weight. Prioritize and ensure signal reading quality.

[0059] By dynamically allocating weights, the underlying speed sampling decision can adaptively match the inventory management needs of the upper layer.

[0060] like Figure 2 As shown, in the warehouse aisle scenario, there is an obstacle directly in front of the robot. The left side of the aisle is a dense area of ​​metal shelves, which is an RFID signal blind spot, while the right side is an open area, which is a high-signal-rate area. In the fan-shaped candidate trajectories generated by this invention, the trajectory biased to the right is due to… The item with the higher score was ultimately selected as the execution trajectory, realizing intelligent collaborative optimization of obstacle avoidance path and business signal quality.

[0061] III. Optimal Instruction Calculation and Issuance

[0062] After completing trajectory sampling and scoring, traverse the dynamic window ( Within a given range of candidate speeds, the optimal speed command with the highest overall score is extracted and sent to the chassis for execution, thus completing trajectory planning and motion execution for a single control cycle.

[0063] During actual execution, the dynamic window is traversed ( The sampling rates of all N groups within the range are extracted based on the Argmax function to make the comprehensive evaluation function. The optimal speed combination to achieve the maximum value: To obtain the optimal linear velocity and optimal angular velocity Then, it is converted into PWM speed control commands that can be directly executed by the underlying motor, driving the robot to perform flexible obstacle avoidance and high-quality inventory operations along a smooth trajectory, avoiding trajectory changes and motor vibration, and extending the service life of the transmission mechanism.

[0064] Combination Figure 1 As shown in the process diagram, the optimal instruction calculation and issuance steps described above are executed cyclically within each control cycle. The system reads the global A* target point, the current robot pose, radar obstacle data, and the warehouse RFID signal confidence heatmap in real time, continuously updating the available dynamic speed window, candidate trajectory scoring results, and optimal output instructions until the robot reaches the global target point, completing the entire warehouse inventory process.

[0065] In practical warehousing applications, this method can complete a single planning cycle in milliseconds, and the output trajectory smoothly conforms to the physical constraints of the motor. At the same time, during obstacle avoidance, it actively moves to areas with high RFID signal strength, avoids signal blind spots, and reduces operating energy consumption. It can simultaneously meet the comprehensive requirements of warehouse inventory robots for real-time performance, safety, and business practicality.

[0066] It should be noted that the above content merely illustrates the technical concept of the present invention and should not be construed as limiting the scope of protection of the present invention. For those skilled in the art, various improvements and modifications can be made without departing from the principle of the present invention, and all such improvements and modifications fall within the scope of protection of the claims of the present invention.

Claims

1. A warehouse inventory robot dynamic window planning method fusing RFID signal perception, characterized in that, It includes the following steps: S1: First, establish the differential drive kinematic model of the robot to understand the robot's response to a given linear velocity ( ) and angular velocity ( In the case of ), how does the pose change recursively? S2: Combining the physical limits and safety constraints of robot motors, construct a set of speed limits. Dynamic acceleration limit set and safety braking limit set The velocity sampling dynamic window formed by taking the intersection of these three sets ( Its mathematical expression is: ； S3: In this dynamic window ( In ) for linear velocity ( ) and angular velocity ( Discrete sampling is performed to generate multiple sets of candidate velocity commands. The trajectory is simulated and predicted for each set of candidate velocities. Finally, a five-dimensional comprehensive evaluation function that includes azimuth angle, safety distance, velocity, RFID signal gain, and energy consumption is used. Each predicted trajectory is scored; S4: Finally, select the five-dimensional comprehensive evaluation function. The candidate speed command with the maximum value is converted into a motor control signal and sent to the robot chassis for execution.

2. The method according to claim 1, characterized in that, The speed limit set mentioned in step S2 ( The parameters are determined by the robot's hardware, and their mathematical expression is: ； in, This is the robot's maximum linear velocity. This represents the robot's maximum angular velocity.

3. The method according to claim 1, characterized in that, The dynamic acceleration limit set mentioned in step S2 ( Limited by the acceleration and deceleration capabilities of the motor, its mathematical expression is: ； in, and These are the linear velocity and angular velocity at the current moment, respectively. and These are the maximum linear acceleration and the maximum linear deceleration, respectively. and These are the maximum angular acceleration and the maximum angular deceleration, respectively. It's about controlling the cycle.

4. The method according to claim 1, characterized in that, The set of safety braking limits mentioned in step S2 ( The core principle is to ensure that the robot can stop in front of an obstacle, and its mathematical expression is: in, For maximum braking deceleration, This refers to the distance to the nearest obstacle measured by the lidar.

5. The method according to claim 1, characterized in that, The five-dimensional comprehensive evaluation function mentioned in step S3 The specific expression is: ； in, , , , , These are the coefficients for each weight. This is the smoothing normalization constant; It is the azimuth evaluation item. It is a safe distance evaluation item. It is a speed evaluation item. This is an RFID signal gain prediction term. It is an energy consumption penalty item.

6. The method according to claim 5, characterized in that, The RFID signal gain prediction item It is obtained by querying a preset RFID signal confidence grid map, and its calculation formula is: ； in, The robot is ranked by candidate speed. Driving to the predicted time window ( Predicted position coordinates at the end of the process. This is the expected signal strength at that location. and These are the maximum and minimum values ​​of the historical statistical signal strength, respectively.

7. The method according to claim 5, characterized in that, Energy consumption penalty Defined as: ； in, For candidate angular velocities, This represents the robot's maximum angular velocity.

8. The method according to claim 5, characterized in that, An adaptive weight adjustment step was also added: when the robot's remaining battery power is detected to be lower than a preset threshold, the weight of the energy consumption penalty term is adjusted. Increase the weight of the RFID signal gain term when the robot enters the preset high-value inventory area. Increase the size to better meet actual operational needs.

9. The method according to claim 5, characterized in that, Safe distance evaluation items A veto mechanism has also been added: if the distance between the predicted trajectory and the nearest obstacle is greater than the preset safety radius... If the value is small, then directly assign the total score to the candidate speed. Setting it to zero is equivalent to directly excluding this candidate speed.

Images