Energy saving optimization method for warehouse equipment based on dynamic task scheduling

Abstract

The present application belongs to the technical field of automatic warehouse, and relates to a warehouse equipment energy-saving optimization method based on dynamic task scheduling, comprising the following steps: initializing an independent potential energy field copy containing a dynamic increment for each candidate equipment; calculating a global incremental energy consumption estimation value and injecting it as a potential energy intensity into a corresponding space-time dimension to generate a quantized potential energy field copy containing a dynamic increment; determining a selected target equipment, an optimal path and a selected causal potential energy field; extracting incremental potential energy data in the selected causal potential energy field and merging it into a global main causal potential energy field to generate an updated main causal potential energy field; and generating a dependent segmented speed instruction sequence according to the optimal path and the selected causal potential energy field and issuing it to the selected target equipment. The present application solves the problem that the existing warehouse scheduling method fails to fully estimate and quantify the chain effect of a single scheduling decision on the overall energy consumption of the equipment cluster when assigning tasks, resulting in a global energy utilization rate that needs to be improved.

Description

Technical Field

[0001] This invention belongs to the technical field of automated warehousing and relates to an energy-saving optimization method for warehousing equipment based on dynamic task scheduling. Background Technology

[0002] In modern automated warehousing systems, numerous autonomous mobile devices, such as autonomous mobile robots (AMRs) or automated guided vehicles (AGVs), work collaboratively to handle, sort, and store goods. To maintain system continuity and throughput, a central scheduling system needs to assign newly generated tasks to appropriate devices in real time. Minimizing overall energy consumption is a core challenge in current warehouse operations. Equipment consumes electricity during movement, acceleration, deceleration, and waiting. These behaviors are interconnected in an environment with multiple devices; the path selection and speed planning of a single device can directly or indirectly affect the movement of other devices, triggering a chain reaction of energy consumption.

[0003] Currently, the solutions commonly used in the industry are usually based on relatively simplified decision-making rules. A common approach is "proximity assignment," where the system assigns a new task to the idle device closest to the task's starting point. This method is logically simple and has a fast response time. A more advanced approach considers more factors, such as incorporating path planning to select the device with the shortest path to execute the task. More complex systems may introduce a short-term traffic manager to resolve impending local path conflicts between devices, avoiding collisions by having one device wait or take a small detour. These methods can maintain basic system operation in simple scenarios with low device density and low task load.

[0004] The aforementioned traditional methods have limitations when dealing with complex, high-density work environments. Scheduling logic based on "nearest assignment" or "shortest path" is essentially local optimization, failing to consider the cascading effects of a single assignment decision on the entire equipment cluster. For example, assigning equipment A with the shortest path might force it to cross a busy intersection, compelling multiple other equipment B, C, and D to decelerate, wait, and then accelerate again. The additional energy consumption generated by this chain reaction is far higher than the total energy consumption of equipment E, which has a slightly longer path but travels in an open area. Existing reactive conflict resolution methods, while ensuring safety, often result in numerous sudden stops and accelerations due to their "last-minute" approach, which are major sources of energy consumption. Current technologies lack a mechanism for proactively and quantitatively assessing the systemic impact of a single decision on total global energy consumption during task assignment.

[0005] Based on the above problems, this invention addresses the issue that existing warehouse scheduling methods fail to fully estimate and quantify the chain reaction of a single scheduling decision on the overall energy consumption of the equipment cluster when assigning tasks, resulting in the need to improve the overall energy utilization rate. Summary of the Invention

[0006] To address the aforementioned problems, this invention provides an energy-saving optimization method for warehousing equipment based on dynamic task scheduling.

[0007] The energy-saving optimization method for warehousing equipment based on dynamic task scheduling includes the following steps: S1. Obtain the real-time status data of the task to be assigned and the candidate devices, and initialize an independent copy of the potential energy field containing dynamic increments for each candidate device based on the preset basic potential energy field. S2. In each potential energy field replica containing dynamic increments, the global device behavior disturbance caused by the deduction decision assumption is calculated, the global incremental energy consumption estimate is injected as the potential energy intensity into the corresponding spatiotemporal dimension, and a quantified potential energy field replica containing dynamic increments is generated. S3. Based on the quantized potential field replica containing dynamic increments, perform conflict arbitration based on potential field cost to determine the selected target device, optimal path, and selected causal potential field. S4. Extract incremental potential energy data from the selected causal potential energy field and merge it into the global main causal potential energy field to generate the updated main causal potential energy field. S5. Generate a segmented speed command sequence based on the optimal path and the selected causal potential energy field, and send it to the selected target device.

[0008] A further aspect of the present invention involves obtaining real-time status data of the task to be assigned and candidate devices, including the following steps: Obtain the running status tags, real-time location coordinates, and current task target point coordinates of all registered devices; Calculate the Euclidean distance between the current position of the device whose running status label is "Task Execution" and the current task target point; Devices with an idle running status label and devices whose Euclidean distance is less than a preset imminent completion distance threshold are defined as candidate devices; The set of candidate devices and their status information are defined as candidate device real-time status data.

[0009] A further aspect of the present invention involves initializing an independent copy of the potential energy field containing dynamic increments for each candidate device, comprising the following steps: Load a two-dimensional matrix representing the passage cost of the static physical layout of the warehouse from persistent storage, and define it as the basic potential field; For each candidate device, a deep copy operation is performed in memory to copy the data of the underlying potential energy field to create an independent copy of the matrix; The matrix copy is defined as a potential energy field copy with dynamic increments specific to the candidate device.

[0010] A further aspect of the present invention addresses the global device behavior perturbation caused by the deduced decision assumptions in each replica of the potential energy field containing dynamic increments, comprising the following steps: Plan and predict trajectories for the corresponding candidate devices in a replica of the potential energy field containing dynamic increments; Simulate the candidate device's journey along the predicted trajectory using discrete time steps; Detect whether the spatial safety envelope of the candidate device overlaps with the planned path of a non-candidate device at future moments; If an overlap occurs, the avoidance or waiting action to be performed by the non-candidate device is determined according to the passage priority rules and is defined as motion adjustment.

[0011] A further aspect of this invention involves calculating the estimated global incremental energy consumption, including the following steps: For each non-candidate device that requires motion adjustment, obtain its instantaneous power function during deceleration, waiting, and re-acceleration phases; Calculate the difference between the instantaneous power function and the device's reference constant-speed cruise power; Integrate the difference over the adjustment period to obtain the incremental energy consumption value of a single disturbance; The incremental energy consumption value is defined as potential energy intensity.

[0012] A further aspect of the present invention involves generating a quantized copy of the potential energy field containing dynamic increments, comprising the following steps: Calculate the estimated energy consumption of the candidate device as it travels along the predicted trajectory; The estimated energy consumption of the device itself is added to the potential energy intensity generated by all affected non-candidate devices to obtain the estimated global incremental energy consumption. The potential energy intensity is superimposed onto the grid value corresponding to the location of the collision in the potential energy field replica containing dynamic increments, generating a quantized potential energy field replica containing dynamic increments.

[0013] A further aspect of the present invention involves performing conflict arbitration based on the potential energy field cost using a quantized copy of the potential energy field containing dynamic increments, comprising the following steps: For each candidate device, the path potential energy integral is calculated along its predicted trajectory in its dedicated, quantized copy of the potential energy field containing dynamic increments. Compare the path potential energy integrals of all candidate devices, and determine the candidate device with the smallest integral value as the selected target device; The predicted trajectory of the selected target device is determined as the optimal path; The quantized potential field copy containing dynamic increments corresponding to the selected target device is determined as the selected causal potential field.

[0014] A further aspect of the present invention generates an updated principal-causal potential field, comprising the following steps: Calculate the difference between the selected causal potential energy field matrix and the basic potential energy field matrix to obtain the incremental potential energy data matrix containing only the current decision disturbance; Add the incremental potential energy data matrix element by element to the global principal-causal potential energy field matrix currently maintained by the system; The result of the addition is persistently stored and defined as the updated principal causal potential field; Release the memory resources occupied by the quantized potential field replicas containing dynamic increments corresponding to all unselected candidate devices.

[0015] A further aspect of the present invention involves generating a dependent segmented velocity command sequence based on the optimal path and the selected causal potential energy field, comprising the following steps: Analyze the optimal path to obtain a continuous sequence of path coordinate points; Traverse the sequence of path coordinate points and read the corresponding potential energy value from the selected causal potential energy field using the coordinate point as the index; The potential energy value is converted into a speed command value through a preset potential energy-velocity mapping function, where the potential energy value and the speed command value are inversely proportional. Each coordinate point is combined with its corresponding speed command value to generate a dependent segmented speed command sequence.

[0016] A further aspect of the present invention includes a preset potential energy-velocity mapping function, comprising the following steps: Set the maximum and minimum permissible driving speeds for the equipment; Set a positive coefficient to adjust the response sensitivity; Calculate the difference between the current potential energy value and the baseline potential energy value; The speed range is scaled by the product of the difference and the positive coefficient, so that when the potential energy value is higher than the basic potential energy reference value, the output speed command value decreases non-linearly from the maximum driving speed and tends to the minimum driving speed.

[0017] In summary, the present invention has the following beneficial technical effects: 1. By creating an independent copy of the potential energy field containing dynamic increments for each candidate device, and independently extrapolating the global device behavior perturbations caused by the decision assumptions within it, the system can proactively quantify the systemic impact of a single scheduling decision on the cluster. The system simulates the additional acceleration, deceleration, and detours of all other devices due to avoidance, waiting, and other behaviors after assigning a candidate device, and calculates the estimated incremental energy consumption of these perturbations. Finally, it combines these indirect energy costs with the direct energy consumption of the candidate device itself.

[0018] 2. By permanently merging the incremental potential energy data corresponding to selected decisions into a global principal-causal potential energy field, a closed-loop feedback and long-term learning mechanism for decision consequences is established. This global principal-causal potential energy field can continuously accumulate traffic disturbance information caused by all historical scheduling decisions. That is, at physical locations where conflicts frequently occur, causing equipment to slow down and wait, the potential energy value will increase over time. This mechanism makes the potential energy field a "memory" carrier of the dynamic changes in the warehouse traffic environment. In subsequent new task scheduling, the path planning algorithm will naturally regard these high-potential-energy areas as high-cost areas and avoid them, achieving adaptive identification and avoidance of congestion hotspots. This reduces reliance on specific bottleneck sections, promotes a balanced distribution of traffic flow throughout the warehouse, and reduces energy waste caused by long-term, structural congestion.

[0019] 3. Through a pre-defined potential energy-velocity mapping function, the system can directly convert the potential energy value at each location along the path into specific, physically executable speed commands, with high potential energy corresponding to low speed. This mechanism ensures that energy-saving behaviors such as avoidance and deceleration predicted during the decision-making phase can be reproduced in the physical world. The equipment will smoothly and proactively adjust its speed according to the commands to cope with potential interactions, rather than relying on onboard sensors for reactive emergency braking. This approach avoids the inrush current and energy loss caused by sudden stops and starts, ensuring that the energy-saving effects calculated during the simulation phase can be realized in actual operation. Attached Figure Description

[0020] To more clearly illustrate the technical solutions in the embodiments of the present invention or the prior art, the accompanying drawings used in the description of the embodiments or the prior art will be briefly introduced below. The drawings are used to provide a further understanding of the present invention.

[0021] Figure 1 A flowchart illustrating an embodiment of this application is disclosed.

[0022] Figure 2 Structural schematic diagrams of embodiments of this application are disclosed. Detailed Implementation

[0023] The following is in conjunction with the appendix Figure 1 - Figure 2A preferred description of the present invention is provided below.

[0024] See attached document Figure 1 This invention proposes an energy-saving optimization method for warehousing equipment based on dynamic task scheduling, comprising the following steps: S1. Obtain the real-time status data of the task to be assigned and the candidate devices, and initialize an independent copy of the potential energy field containing dynamic increments for each candidate device based on the preset basic potential energy field. S2. In each potential energy field replica containing dynamic increments, the global device behavior disturbance caused by the deduction decision assumption is calculated, the global incremental energy consumption estimate is injected as the potential energy intensity into the corresponding spatiotemporal dimension, and a quantified potential energy field replica containing dynamic increments is generated. S3. Based on the quantized potential field replica containing dynamic increments, perform conflict arbitration based on potential field cost to determine the selected target device, optimal path, and selected causal potential field. S4. Extract incremental potential energy data from the selected causal potential energy field and merge it into the global main causal potential energy field to generate the updated main causal potential energy field. S5. Generate a segmented speed command sequence based on the optimal path and the selected causal potential energy field, and send it to the selected target device.

[0025] In one embodiment of the present invention, step S1 includes the following steps: The system acquires real-time status data of the tasks to be assigned and all candidate devices, and for each candidate device, initializes a dynamically incrementing copy of the potential energy field representing the decision hypothesis of "assigning this device". When the warehouse system generates a new task to be assigned, it acquires in real-time the current location, speed, battery level, and planned path information of all candidate devices that are idle or about to complete their tasks. For each candidate device, a dedicated, temporary copy of the potential energy field containing dynamic increments is created. This superposition state initially contains only the basic potential energy value corresponding to the static physical layout of the warehouse, serving as the basis for subsequent causal chain deduction.

[0026] Specifically, step S1 is executed by a central scheduling system deployed on a cloud server or edge computing node. This process begins with the Warehouse Management System (WMS) publishing a data packet containing a unique task identifier, pickup point coordinates, and storage point coordinates via an internal message queue, such as using the AMQP protocol. The central scheduling system subscribes to this message queue and, upon receiving the data packet, triggers the execution logic for this step. It sends a broadcast data query request to the equipment management subsystem or retrieves the status of all registered automated warehousing equipment through a polling mechanism. For each piece of equipment, the system obtains its equipment ID, operating status label, real-time location coordinates, vector velocity, remaining battery percentage, and the planned path sequence if it is currently executing a task.

[0027] The system iterates through the status information of all devices, filtering out candidate devices based on either a device's "idle" status label or a "task in progress" status label where the Euclidean distance between its current position and the current task target point is less than a preset near-completion distance threshold. After filtering, the system obtains a set of N candidate devices and complete real-time status data for each device, defining this data set as the candidate device real-time status data. For each of these N candidate devices, the system allocates and initializes an independent, temporary data structure in memory, namely a copy of the potential energy field containing dynamic increments.

[0028] Specifically, the initialization process involves the system loading a pre-calculated basic potential energy field representing the static physical layout of the warehouse from persistent storage, such as a Redis database or a local file. This basic potential energy field is a two-dimensional matrix, with its dimensions corresponding to the rasterized representation of the warehouse map. The value of each element in the matrix represents the fixed access cost of the corresponding physical location. For the i-th candidate device, the system performs a deep copy operation, completely copying the data of the entire basic potential energy field to a new memory address space, thereby generating a matrix that is identical to the initial content of the basic potential energy field but independent of it. This matrix is ​​defined as the potential energy field replica containing dynamic increments corresponding to the i-th candidate device. After this step, the system generates a total of N potential energy field replicas containing dynamic increments, each of which serves as the basis for independent causal inference in subsequent step S2.

[0029] The potential energy field replica containing dynamic increments is a temporary data structure representing the future spatiotemporal energy cost distribution under specific decision assumptions. During initialization, it is visualized as a two-dimensional digital matrix, with the matrix index corresponding to the rasterized coordinates of the warehouse, and the matrix value being a dimensionless potential energy value. "Task nearing completion" refers to a device performing a task whose current position is less than or equal to a preset near-completion distance threshold between its current location and the task's target point. This threshold is set empirically, typically between 3 and 5 meters. This range is based on the assumption that the device's average travel speed is 1.5 meters per second, and the system requires approximately 1-2 seconds to complete a full decision calculation. Setting a threshold of 3-5 meters ensures the system has sufficient time to plan subsequent tasks for the device, while avoiding premature task assignment that could cause the device to wait in place.

[0030] The base potential value is pre-calculated and stored offline to represent the passage cost of the warehouse static environment. The setting rule is that for grids in open areas, the base potential value is set to each low baseline value, such as 1.0. Candidate devices refer to autonomous mobile devices whose status is "idle" or "task about to be completed" when a new task is received.

[0031] For grids occupied by permanent obstacles such as walls and shelf columns, their base potential energy value is set to a maximum or infinity, and crossing can be prohibited in path planning; for known congestion-prone areas such as narrow passages and frequent intersections, the base potential energy value of each height is set according to its historical congestion statistics or physical width, for example, 5.0 to 15.0.

[0032] For example, suppose the Warehouse Management System (WMS) generates a new task T101 with a pickup point at coordinates (20, 80) and a storage point at (90, 15). After receiving this task, the central dispatch system queries the status of all three devices AGV-01, AGV-02, and AGV-03 in the system. The real-time status data obtained is as follows: AGV-01 is in "idle" status, located at (22, 75), and has a battery level of 88%; AGV-02 is in "task in progress," with the target point of the current task T099 at (50, 50), its current location at (48, 51), and its battery level at 65%; AGV-03 is in "task in progress," its current location at (10, 10), and its target point is far away.

[0033] The system begins screening candidate devices. AGV-01 is directly selected as a candidate device because its status is "idle". The system calculates the distance between AGV-02 and its task target point, which is approximately 2.24 meters (sqrt((50-48)^2+(50-51)^2)). Assuming the system's preset near-completion distance threshold is 3 meters, since 2.24 meters is less than 3 meters, AGV-02 is determined to be "task nearing completion" and is also selected as a candidate device. AGV-03 is not selected because it is too far from the target point.

[0034] The system identifies a set of candidate devices including AGV-01 and AGV-02 and obtains their real-time status data. A 100x100 two-dimensional matrix is ​​loaded from the database as the basic potential energy field. For example, the value at (25, 60) represents a narrow passage with a value of 10.0, while the value for most open areas is 1.0. The system then creates a 100x100 matrix named "Potential Energy Field Copy_1 with Dynamic Increments" for AGV-01, setting all its elements to be identical to the basic potential energy field. Simultaneously, the system creates another independent 100x100 matrix named "Potential Energy Field Copy_2 with Dynamic Increments" for AGV-02, also performing a deep copy operation to ensure its initial content is consistent with the basic potential energy field. At this point, step S1 is complete, generating two candidate devices and their status data, as well as two independent, identical copies of the potential energy field containing dynamic increments, which are then output as a whole to subsequent steps.

[0035] In one embodiment of the present invention, step S2 includes the following steps: In each potential energy field replica containing dynamic increments, the global device behavior disturbance caused by the corresponding decision assumptions is independently deduced, and the resulting global incremental energy consumption estimate is quantified and injected into the corresponding spatiotemporal dimension of the superposition state to generate a quantized potential energy field replica containing dynamic increments.

[0036] Within each replica of the potential energy field containing dynamic increments, the simulation assigns its corresponding candidate device to perform the assigned task and predicts the trajectory of that candidate device. Based on the predicted trajectory of the candidate device, the speed and path adjustments that all other devices must make to avoid, wait for, or follow that candidate device are deduced, forming a causal chain of decisions.

[0037] In the calculation of the causal chain of decision-making, the incremental energy consumption prediction of all devices due to additional acceleration, deceleration, detour, and waiting, including the candidate device itself and other affected devices, is used as the potential energy intensity. This energy consumption value is injected into the potential energy field replica containing dynamic increments and the spatiotemporal coordinates corresponding to the predicted location and time of the disturbance event, generating a quantized potential energy field replica containing dynamic increments.

[0038] Specifically, this step takes the real-time status data of the candidate devices generated in step S1 above and N independent potential energy field replicas containing dynamic increments as input. The central scheduling system starts an independent parallel computing thread or logic processing unit for each candidate device to deduce decision assumptions in its dedicated potential energy field replica containing dynamic increments. In the deduction process for the i-th candidate device, the system uses the candidate device's current location, task pickup point coordinates, and task storage point coordinates, and its dedicated potential energy field replica containing dynamic increments as a travel cost map, to calculate the path with the minimum potential energy integral using the A* algorithm or Dijkstra's algorithm. This path is defined as the predicted trajectory of the candidate device, and the system enters the decision causal chain deduction stage.

[0039] The system simulates the candidate device's movement along its predicted trajectory in discrete time steps, such as 0.1 seconds. At each time step, the system checks whether the candidate device's spatial safety envelope overlaps with the spatial safety envelopes of all other non-candidate devices' planned paths at the same time. If an overlap or conflict is detected at a future time t with a location (x, y), the system determines that one of the devices needs to adjust its movement according to a preset traffic priority rule. Typically, the candidate device executing the newly assigned task has a lower priority; therefore, the other affected device is forced to perform a movement adjustment to avoid or wait. This adjustment involves the system replanning its speed curve, which includes deceleration, waiting, and re-acceleration. Subsequently, the system calculates the incremental energy consumption caused by this movement adjustment. This incremental energy consumption is calculated based on the difference between the power consumption of the affected device during deceleration, waiting, and re-acceleration phases and the baseline constant-speed driving power consumption, integrated over time.

[0040] The system accumulates the incremental energy consumption generated by all identified motion adjustments performed by other devices in the decision causal chain, and adds this to the total energy consumption generated by the candidate device itself executing its predicted trajectory, to obtain the total energy value, which is defined as the global incremental energy consumption estimate. The system uses the incremental energy consumption calculated for each motion adjustment as a potential energy intensity value, incrementally adding it to its corresponding potential energy field replica containing dynamic increments. Specifically, this incremental energy consumption value is added to the existing potential energy value on the grid at the conflict location (x, y). After completing the calculation and potential energy injection for all potential conflicts, the candidate device's dedicated potential energy field replica containing dynamic increments is updated to a quantized potential energy field replica containing dynamic increments. This complete process is performed independently in the deductions corresponding to all candidate devices, generating N quantized potential energy field replicas containing dynamic increments and N corresponding global incremental energy consumption estimates.

[0041] In the decision causal chain, for the decision assumption of the i-th candidate device, the corresponding global incremental energy consumption prediction value is... Calculated using the following formula:

[0042] Where E_global,i is the estimated global incremental energy consumption caused by assigning a task to the i-th candidate device. E_candidate,i is the estimated energy consumed by the i-th candidate device itself to complete the task path, which can be calculated from its total path length, average speed, and rated power. M is the total number of other devices that are disturbed by the movement of the i-th candidate device.

[0043] E_disturbance,j represents the incremental energy consumption of the j-th affected device due to motion adjustment, and it is calculated as follows:

[0044] Where t_adj is the total duration of the entire adjustment process; P_adj(t) is the instantaneous power of the affected device during the adjustment process, which has different functional forms during deceleration, waiting, and acceleration phases. P_cruise is the reference power of the device when cruising at a normal constant speed without interference. The actual incremental energy consumption is obtained by integrating the power difference over the entire adjustment time.

[0045] In this context, the decision causal chain refers to the simulated motion trajectory of a candidate device as the initial cause, triggering a series of path or speed adjustments that other devices must make to avoid conflict. These adjustments constitute a series of interconnected events. The global incremental energy consumption forecast is a quantitative indicator used to measure the net impact of adopting a certain decision hypothesis on the total energy consumption of the entire warehouse equipment cluster over future time. Potential energy intensity is numerically equal to the incremental energy consumption forecast generated by a single avoidance, waiting, or other disturbance event, and its unit is joules (J) or equivalent energy units. This value is directly superimposed on the potential energy field matrix as a dimensionless cost.

[0046] The power parameters in the energy consumption calculation model are set based on equipment calibration experiments. For example, for an autonomous mobile robot with a rated load of 500kg, its idle waiting power P_wait is 50W, its cruising power P_cruise at a constant speed of 1.5m / s is 200W, and its peak power P_accel when accelerating at an acceleration of 0.5m / s^2 is 350W. Energy recovery is usually achieved through motor braking during deceleration. To simplify the model, its equivalent power consumption can be set to 50W-100W.

[0047] For example, the system has two independent simulation tasks. For AGV-01, in its dedicated potential energy field copy_1 containing dynamic increments, its predicted trajectory from the current position (22, 75) to the pickup point (20, 80) and then to the storage point (90, 15) is planned. When simulating this trajectory, the system finds that when AGV-01 travels to (55, 65), it will conflict with the path of AGV-03, which is performing a task. According to the priority rule, AGV-03 needs to perform avoidance, and its movement is adjusted to decelerate for 1 second, wait for 3 seconds, and accelerate for 1 second. Based on the preset power, the system calculates the incremental energy consumption E_disturbance_AGV-03 caused by this disturbance as (350W-200W)×1s+(50W-200W)×3s+(350W-200W)×1s, which is a net incremental energy consumption of 300 joules.

[0048] AGV-01 travels a path of approximately 100.5 meters, with an estimated energy consumption of 13400 joules (E_candidate_1). The estimated global incremental energy consumption for AGV-01 is 13400J + 300J = 13700J. A value of 300 is injected into the potential energy field replica_1 containing dynamic increments, changing its potential energy value at coordinates (55, 65) from the initial 1.0 to 301.0. This modified field is the quantized potential energy field replica_1 containing dynamic increments. For AGV-02, the system plans its predicted trajectory from (48, 51) to complete the task in its dedicated potential energy field replica_2 containing dynamic increments. Simulations show that its path has no spatiotemporal intersection with the path of AGV-03, therefore the number of disturbed devices is 0, and the incremental energy consumption is 0 joules. The AGV-02 travels a total distance of approximately 135.8 meters, with an estimated energy consumption of 18,100 joules (E_candidate_2).

[0049] The estimated global incremental energy consumption for AGV-02 is 18100J + 0J = 18100J. Since there are no disturbance events, the potential energy field replica_2 containing the dynamic increment has no potential energy injection, and its content remains consistent with the basic potential energy field. This field is the quantized potential energy field replica_2 containing the dynamic increment. Step S2 outputs two quantized potential energy field replicas containing the dynamic increment to the next step, along with their corresponding estimated global incremental energy consumption values ​​of 13700J and 18100J.

[0050] In one embodiment of the present invention, step S3 includes the following steps: Based on the quantized copies of the potential energy field containing dynamic increments, conflict arbitration based on the cost of the potential energy field is performed to determine the unique selected target device and its optimal path and the selected causal potential energy field.

[0051] For each candidate device, within its dedicated quantized potential energy field replica containing dynamic increments, the path potential energy integral from the candidate device's current position to the mission target point is calculated. This integral value is numerically equivalent to the global incremental energy consumption estimate calculated in step S2. The path potential energy integral values ​​calculated by all candidate devices are compared, and the candidate device with the smallest integral value is determined as the selected target device. The path corresponding to the selected target device is determined as the optimal path, and its dedicated quantized potential energy field replica containing dynamic increments is determined as the selected causal potential energy field.

[0052] Specifically, the system receives the quantized potential energy field replica containing dynamic increments generated in step S2 and the corresponding global incremental energy consumption estimate as input. The central scheduling system performs conflict arbitration based on potential energy field cost for each candidate device. For the i-th candidate device, the system calculates the potential energy integral of the path along the predicted trajectory planned for it in step S2 within its dedicated quantized potential energy field replica containing dynamic increments. This calculation process involves discretizing the predicted trajectory to obtain a sequence of path points, and then accumulating the potential energy value corresponding to the grid where each path point is located. The accumulated result is the path potential energy integral, which is designed to be completely equal to the global incremental energy consumption estimate calculated in step S2, forming a causal interlock between decision inference and cost accounting.

[0053] After calculating the path potential energy integral for all N candidate devices, the system enters the comparison and decision-making stage. It compares all N path potential energy integral values ​​and determines the minimum value. The candidate device with the minimum path potential energy integral value is uniquely identified as the selected target device. The system formally adopts the predicted trajectory used by the selected target device in step S2 as the optimal path. Simultaneously, it solidifies the quantized potential field copy corresponding to the selected target device, containing dynamic increments, into the selected causal potential field. Finally, it outputs three decision results: the selected target device, the optimal path, and the selected causal potential field, for use in subsequent steps and environment updates.

[0054] The formula explains that for the i-th candidate device, the path potential energy integral C_i is calculated as follows:

[0055] Where P_i is the predicted trajectory generated by the i-th candidate device in step S2. V_i(p) is the potential energy value at position vector p in the quantized copy of the potential energy field containing dynamic increments specific to the i-th candidate device. This integral is represented in the discretized implementation as the summation of the potential energy values ​​of all grids along the path. According to the design of the present invention, this integral value is numerically equivalent to the global incremental energy consumption prediction value E_global,i calculated in step S2, i.e., C_i=E_global,i / k, where k is the energy-to-potential energy conversion coefficient, assuming k=1, and the unit is joules / potential energy unit.

[0056] The final selected target device index k satisfies the following condition:

[0057] The path potential energy integral is the final measure of the total cost incurred by a candidate device traveling along its predicted path. This cost includes not only its own energy consumption but also all indirect energy consumption resulting from its environmental disturbances. The selected target device is the one whose decision assumptions have been proven to have the least impact on the future energy consumption of the entire system among all candidates. The optimal path is not necessarily the path with the shortest physical distance, but rather the path with the smallest estimated global incremental energy consumption. The selected causal potential energy field is a distribution map of the future spatiotemporal energy consumption cost corresponding to the ultimately adopted decision, reflecting the objective impact of this scheduling decision on the warehousing environment.

[0058] For example, the central scheduling system receives two quantized copies of the potential energy field containing dynamic increments, along with corresponding global incremental energy consumption estimates of 13700J and 18100J. For AGV-01, the system calculates the path potential energy integral along its planned predicted trajectory in its dedicated quantized copy of the potential energy field containing dynamic increments, _1. This calculation combines the cost of the basic path with the 300-point potential energy representing the disturbance energy consumption injected at coordinates (55, 65), ultimately yielding a path potential energy integral of 13700.

[0059] For AGV-02, the system calculates the path potential energy integral in its dedicated quantized potential energy field copy_2 containing dynamic increments. Since its path does not cause any disturbance, the integral result equals its own energy consumption estimate of 18100. The system compares these two integral values; since 13700 is less than 18100, the system makes the final decision. The system identifies AGV-01 as the selected target device. The system determines the planned path of AGV-01 from (22, 75) through (20, 80) to (90, 15) as the optimal path. The system identifies the quantized potential energy field copy_1 containing incremental potential energy at coordinates (55, 65) as the selected causal potential energy field. These three data points—the selected target device AGV-01, the optimal path data, and the selected causal potential energy field matrix—are packaged as the output of this step.

[0060] In one embodiment of the present invention, step S4 includes the following steps: Based on the selected causal potential field, the impact of this scheduling decision on the global environment is solidified, generating an updated master causal potential field to guide all subsequent scheduling decisions. The incremental potential data triggered by this decision, contained in the selected causal potential field, is incrementally merged into the system-maintained global master causal potential field. All quantized copies of potential fields containing dynamic increments corresponding to other unselected candidate devices are discarded to ensure the uniqueness of the decision and the continuity of environmental evolution. The merged global master causal potential field is used as the updated master causal potential field, serving as the decision basis for step S1 in the next new task assignment.

[0061] Specifically, this step is triggered after step S3, which determines the selected target device, optimal path, and selected causal potential field. The central scheduling system first performs a potential energy solidification operation. The system loads the current version of the global master-causal potential field from persistent storage, such as a memory database or a file on a solid-state drive. This field is the cumulative result of the effects of all previously executed scheduling decisions. The system retrieves the basic potential field used in step S1 of this scheduling process. By performing a matrix subtraction operation, the system calculates the difference between the selected causal potential field and the basic potential field, obtaining a difference matrix that only contains the potential energy increment caused by this decision. This matrix is ​​the incremental potential energy data. The system performs a matrix addition operation, adding the incremental potential energy data element-wise to the loaded global master-causal potential field to generate a merged potential field. The merged potential field is written to persistent storage by the system, overwriting the old version of the global master-causal potential field, and is defined as the updated master-causal potential field.

[0062] After the merge is complete, the system performs a resource cleanup operation, iterating through all candidate devices that were not selected in this decision-making process and releasing the memory resources they occupied during the deduction process in step S2. This involves discarding all non-selected, quantized copies of the potential energy field containing dynamic increments. This ensures the uniqueness of the decision and prevents temporary deduction data from causing continuous memory consumption. The updated principal-causal potential energy field will be loaded as the base potential energy field when a new task is assigned, incorporating the long-term environmental impact of this decision into future decision-making considerations, thus forming a closed-loop evolution of the decision-making process.

[0063] The update process of the global principal-causal potential field can be represented by the following equation:

[0064] in, , It is the updated principal causal potential field matrix generated after executing this scheduling decision. It is the global principal-causal potential energy field matrix maintained by the system before this scheduling decision. It is the selected causal potential field matrix determined in step S3. This is the basic potential field matrix used for initialization in step S1 of this scheduling process, and in its implementation... yes A complete copy at the start of scheduling. This is the incremental potential energy data matrix, whose non-zero elements only exist at the spatiotemporal location of the disturbance event triggered by this decision.

[0065] Incremental potential energy data refers to the set of incremental energy consumption estimates generated by the movement adjustments of all other equipment triggered by the predicted behavior of the selected target equipment during the simulation of this scheduling decision. These energy consumption values ​​are fixed in the corresponding positions of the potential energy field matrix. The global principal-causal potential energy field is a persistent data structure that is continuously updated throughout the entire system lifecycle. It records the cumulative causal impact of all historical scheduling decisions on the warehouse traffic environment and serves as the foundation for the system's long-term learning and adaptive optimization. The updated principal-causal potential energy field represents the latest state of the global principal-causal potential energy field after completing this incremental merging operation.

[0066] For example, the input is the selected causal potential energy field, i.e., the quantized copy of the potential energy field containing dynamic increments corresponding to AGV-01. The central scheduling system loads the current global master causal potential energy field. Assuming that no disturbance events have occurred before this task T101, the content of this field is exactly the same as the initial static physical layout of the warehouse, i.e., the basic potential energy field, and its potential energy value at coordinates (55, 65) is 1.0. The system calculates the incremental potential energy data by subtracting the basic potential energy field matrix used in step S1 from the selected causal potential energy field matrix. Since the value of the selected causal potential energy field at coordinates (55, 65) is 301.0, and the value of the basic potential energy field at this point is 1.0, the calculated incremental potential energy data has a value of 300.0 at this point, and a value of 0 at all other locations.

[0067] The system merges the incremental potential energy data into the global principal causal potential energy field. Specifically, it adds the original potential energy value of 1.0 at (55, 65) to the incremental value of 300.0, obtaining a new potential energy value of 301.0. This updated matrix is ​​then saved as the updated principal causal potential energy field. The system discards the quantized potential energy field copy_2 containing the dynamic increment corresponding to the unselected AGV-02, releasing the memory it occupies. Finally, when the next new task, such as T102, is generated, the system will load this latest updated principal causal potential energy field with a potential energy value of 301.0 at (55, 65) as the basis for decision-making in step S1.

[0068] In one embodiment of the present invention, step S5 includes the following steps: Based on the optimal path and the selected causal potential energy field, a dependent and segmented speed command sequence is generated for the selected target device and then executed. The optimal path is parsed to obtain a sequence of continuous coordinate points along the path. For each coordinate point on the path, the corresponding potential energy value is read from the selected causal potential energy field. Using a preset potential energy-velocity mapping function, the read potential energy value is directly converted into a specific speed control command; the higher the potential energy value, the lower the mapped speed command value.

[0069] All speed control commands generated along the optimal path are combined into a complete speed command sequence in chronological order and sent to the selected target equipment for execution, ensuring that the predicted energy-saving effect is realized at the physical level.

[0070] Specifically, this step is executed in a dedicated service module for command generation. This module receives the selected target device, optimal path, and selected causal potential energy field output from step S3 as input. The service module parses the optimal path, converting it from a high-level path description into a sequence of path points composed of continuous coordinates. The system traverses each coordinate point in this path point sequence serially. For the k-th coordinate point in the sequence, the system uses that coordinate as an index to read the corresponding potential energy value V_k from the selected causal potential energy field matrix. Then, the system calls a preset potential energy-velocity mapping function, using the read potential energy value V_k as input, to calculate the specific velocity command value v_k. This mapping function is designed with an inverse relationship; that is, the higher the input potential energy value, the lower the output velocity command value, enabling the device to automatically decelerate in predicted high-risk or high-energy-consumption areas.

[0071] The system combines each coordinate point with its corresponding speed command value v_k into a tuple (coordinate point_k, v_k). After traversing the entire path point sequence, the system combines all generated tuples into a complete, ordered list structure according to their order in the path. This list is the dependent, segmented speed command sequence. This speed command sequence is then sent to the onboard controller of the selected target device via a wireless communication network, for example, using a custom binary format or JSON format data packet based on the TCP protocol. The controller is responsible for executing the speed requirements for each segment of the path.

[0072] The potential energy-velocity mapping function f_map converts the potential energy value V_k into a velocity command v_k. Its specific functional form can be a bounded inverse proportional function.

[0073] Here, v_k is the speed command generated for the k-th point on the path. v_max is the maximum allowed speed of the device, and v_min is the minimum speed at which the device maintains stable motion. V_k is the potential energy value of the corresponding point read from the selected causal potential energy field. V_base is the baseline value of the basic potential energy field when there is no disturbance. α is a positive sensitivity coefficient used to adjust the drastic response of speed to changes in potential energy; the larger the value, the faster the speed decreases as potential energy increases.

[0074] The dependent, segmented speed command sequence is a high-level command format that decomposes the path into a series of small segments and specifies a speed target for each segment directly determined by a selected causal potential energy field, rather than simply providing a single endpoint and average speed. The potential energy-velocity mapping function is a key step in realizing the energy-saving effect of this invention at the physical level, directly transforming the abstract cost prediction in the decision-making stage into physical motion parameters in the equipment execution stage. `v_max` is set according to the equipment hardware specifications and storage safety regulations, for example, 1.5 m / s. `v_min` is set to prevent the equipment from stopping completely unnecessarily, for example, 0.1 m / s.

[0075] V_base is typically set to 1.0, corresponding to the potential energy in an open, undisturbed region. The sensitivity coefficient α is an empirical parameter determined through simulation or field testing, and its value usually ranges from 0.05 to 0.5. If α is set to 0.1, it means that when the potential energy value is 10 points higher than the baseline value, the denominator becomes 2, and the speed will decrease from v_min to half the difference between the maximum and minimum speeds.

[0076] For example, the instruction generation module receives the selected target device AGV-01, its optimal path, and the selected causal potential energy field. The optimal path is parsed to obtain a list containing a series of coordinate points from (22, 75) to (90, 15). The system begins traversing this list. When a regular point in the path is reached, such as (30, 78), the system reads the potential energy value of that point from the selected causal potential energy field as V_k = 1.0. Assume the system's preset parameters are... The system inputs the formula to calculate the velocity at that point. When the system traverses to the perturbation point (55, 65) in the path, it reads the high potential energy value of that point from the selected causal potential energy field as V_k = 301.0. The system then substitutes this value into the formula to calculate the velocity command for this point: .

[0077] In this way, the system generates a corresponding speed value for each point on the path. The system generates a dependent, segmented speed command sequence in the form of [((22, 75), 1.5), ..., ((55, 65), 0.145), ..., ((90, 15), 1.5)]. This sequence is encoded into JSON format and sent to AGV-01 via Wi-Fi. After receiving this sequence, AGV-01's motion controller will smoothly reduce its speed to 0.145 m / s when it travels to the area near (55, 65), physically realizing the avoidance behavior predicted in step S2, and ensuring the final realization of global energy consumption optimization.

[0078] See appendix Figure 2The present invention also proposes an energy-saving optimization system for warehousing equipment based on dynamic task scheduling, comprising the following modules: The status acquisition and potential energy field initialization module acquires real-time status data of the task to be assigned and candidate devices, and initializes an independent copy of the potential energy field containing dynamic increments for each candidate device based on the preset basic potential energy field. The causal inference and energy consumption quantification module infers the global device behavior disturbance caused by the decision assumption in each potential energy field replica containing dynamic increments, calculates the global incremental energy consumption estimate and injects it as potential energy intensity into the corresponding spatiotemporal dimension, and generates a quantified potential energy field replica containing dynamic increments. The interlock decision and path arbitration module performs conflict arbitration based on the potential energy field cost, based on a quantified copy of the potential energy field containing dynamic increments, to determine the selected target device, the optimal path, and the selected causal potential energy field. The main potential field update and solidification module extracts incremental potential energy data from the selected causal potential energy field and merges it into the global main causal potential energy field to generate the updated main causal potential energy field. The speed command sequence generation module generates a dependent segmented speed command sequence based on the optimal path and the selected causal potential energy field, and sends it to the selected target device.

[0079] Each of the modules can be implemented in whole or in part through software, hardware, or a combination thereof. It supports hardware embedded in or independent of the processor in the computer device, and also supports software stored in the memory of the computer device, so that the processor can call and execute the operations corresponding to each of the above modules.

[0080] The above embodiments are only used to illustrate the technical solutions of the present invention, and are not intended to limit it. Although the present invention has been described in detail with reference to the foregoing embodiments, those skilled in the art should understand that modifications can still be made to the technical solutions described in the foregoing embodiments, or equivalent substitutions can be made to some of the technical features. Such modifications or substitutions do not cause the essence of the corresponding technical solutions to deviate from the spirit and scope of the technical solutions of the embodiments of the present invention, and should all be included within the protection scope of the present invention.

Claims

1. A method for energy-saving optimization of warehousing equipment based on dynamic task scheduling, characterized in that, Includes the following steps: S1. Obtain the real-time status data of the task to be assigned and the candidate devices, and initialize an independent copy of the potential energy field containing dynamic increments for each candidate device based on the preset basic potential energy field. S2. In each potential energy field replica containing dynamic increments, the global device behavior disturbance caused by the deduction decision assumption is calculated, the global incremental energy consumption estimate is injected as the potential energy intensity into the corresponding spatiotemporal dimension, and a quantified potential energy field replica containing dynamic increments is generated. S3. Based on the quantized potential field replica containing dynamic increments, perform conflict arbitration based on potential field cost to determine the selected target device, optimal path, and selected causal potential field. S4. Extract incremental potential energy data from the selected causal potential energy field and merge it into the global main causal potential energy field to generate the updated main causal potential energy field. S5. Generate a segmented speed command sequence based on the optimal path and the selected causal potential energy field, and send it to the selected target device.

2. The energy-saving optimization method for warehousing equipment based on dynamic task scheduling according to claim 1, characterized in that, Obtaining real-time status data of tasks to be assigned and candidate devices includes the following steps: Obtain the running status tags, real-time location coordinates, and current task target point coordinates of all registered devices; Calculate the Euclidean distance between the current position of the device whose running status label is "Task Execution" and the current task target point; Devices with an idle running status label and devices whose Euclidean distance is less than a preset imminent completion distance threshold are defined as candidate devices; The set of candidate devices and their status information are defined as candidate device real-time status data.

3. The energy-saving optimization method for warehousing equipment based on dynamic task scheduling according to claim 2, characterized in that, Initialize an independent copy of the potential field containing dynamic increments for each candidate device, including the following steps: Load a two-dimensional matrix representing the passage cost of the static physical layout of the warehouse from persistent storage, and define it as the basic potential field; For each candidate device, a deep copy operation is performed in memory to copy the data of the underlying potential energy field to create an independent copy of the matrix; The matrix copy is defined as a potential energy field copy with dynamic increments specific to the candidate device.

4. The energy-saving optimization method for warehousing equipment based on dynamic task scheduling according to claim 1, characterized in that, The global device behavior perturbation caused by the extrapolation decision assumptions in each replica of the potential energy field containing dynamic increments includes the following steps: Plan and predict trajectories for the corresponding candidate devices in a replica of the potential energy field containing dynamic increments; Simulate the candidate device's journey along the predicted trajectory using discrete time steps; Detect whether the spatial safety envelope of the candidate device overlaps with the planned path of a non-candidate device at future moments; If an overlap occurs, the avoidance or waiting action to be performed by the non-candidate device is determined according to the passage priority rules and is defined as motion adjustment.

5. The energy-saving optimization method for warehousing equipment based on dynamic task scheduling according to claim 4, characterized in that, Calculating the estimated global incremental energy consumption includes the following steps: For each non-candidate device that requires motion adjustment, obtain its instantaneous power function during deceleration, waiting, and re-acceleration phases; Calculate the difference between the instantaneous power function and the device's reference constant-speed cruise power; Integrate the difference over the adjustment period to obtain the incremental energy consumption value of a single disturbance; The incremental energy consumption value is defined as potential energy intensity.

6. The energy-saving optimization method for warehousing equipment based on dynamic task scheduling according to claim 5, characterized in that, Generating a quantized copy of the potential field containing dynamic increments includes the following steps: Calculate the estimated energy consumption of the candidate device as it travels along the predicted trajectory; The estimated energy consumption of the device itself is added to the potential energy intensity generated by all affected non-candidate devices to obtain the estimated global incremental energy consumption. The potential energy intensity is superimposed onto the grid value corresponding to the location of the collision in the potential energy field replica containing dynamic increments, generating a quantized potential energy field replica containing dynamic increments.

7. The energy-saving optimization method for warehousing equipment based on dynamic task scheduling according to claim 1, characterized in that, Conflict arbitration based on potential field cost is performed using a quantized copy of the potential field containing dynamic increments, including the following steps: For each candidate device, the path potential energy integral is calculated along its predicted trajectory in its dedicated, quantized copy of the potential energy field containing dynamic increments. Compare the path potential energy integrals of all candidate devices, and determine the candidate device with the smallest integral value as the selected target device; The predicted trajectory of the selected target device is determined as the optimal path; The quantized potential field copy containing dynamic increments corresponding to the selected target device is determined as the selected causal potential field.

8. The energy-saving optimization method for warehousing equipment based on dynamic task scheduling according to claim 1, characterized in that, Generating the updated principal-causal potential field includes the following steps: Calculate the difference between the selected causal potential energy field matrix and the basic potential energy field matrix to obtain the incremental potential energy data matrix containing only the current decision disturbance; Add the incremental potential energy data matrix element by element to the global principal-causal potential energy field matrix currently maintained by the system; The result of the addition is persistently stored and defined as the updated principal causal potential field; Release the memory resources occupied by the quantized potential field replicas containing dynamic increments corresponding to all unselected candidate devices.

9. The energy-saving optimization method for warehousing equipment based on dynamic task scheduling according to claim 1, characterized in that, Generate a piecewise velocity command sequence based on the optimal path and the selected causal potential energy field, including the following steps: Analyze the optimal path to obtain a continuous sequence of path coordinate points; Traverse the sequence of path coordinate points and read the corresponding potential energy value from the selected causal potential energy field using the coordinate point as the index; The potential energy value is converted into a speed command value through a preset potential energy-velocity mapping function, where the potential energy value and the speed command value are inversely proportional. Each coordinate point is combined with its corresponding speed command value to generate a dependent segmented speed command sequence.

10. The energy-saving optimization method for warehousing equipment based on dynamic task scheduling according to claim 9, characterized in that, The preset potential energy-velocity mapping function includes the following steps: Set the maximum and minimum permissible driving speeds for the equipment; Set a positive coefficient to adjust the response sensitivity; Calculate the difference between the current potential energy value and the baseline potential energy value; The speed range is scaled by the product of the difference and the positive coefficient, so that when the potential energy value is higher than the basic potential energy reference value, the output speed command value decreases non-linearly from the maximum driving speed and tends to the minimum driving speed.

Images