Method and system for task allocation of construction robots in conjunction with charge management

Abstract

The present application relates to the technical field of building robot operation management and job scheduling, and particularly relates to a method and system for building robot task allocation combined with power management, obtaining parameter information, driving information and residual power of a driving section, obtaining a task set to be executed and determining a preset task by calculating and predicting power consumption of a power consumption prediction model; updating the predicted power consumption within a preset time window, determining an entry threshold value / exit threshold value according to a power feasibility judgment sequence to obtain an equivalent flip frequency; setting a task reassignment token bucket replenishment rate and an upper limit according to an invalid energy consumption proportion sequence; allocating and monitoring, reassigning and recalculating if the exit threshold value is lower and the tokens are sufficient, otherwise issuing a charging task, updating energy consumption prediction in real time and selecting the optimal task to reduce misallocation; based on the feasibility flip adaptive entry / exit threshold value, relieving jitter, improving continuity and safety; and feeding back the token bucket according to invalid energy consumption to suppress frequent reassignment, reduce energy consumption and increase efficiency.

Description

Technical Field

[0001] This invention relates to the field of construction robot operation management and job scheduling technology, and more specifically, to a method and system for allocating construction robot tasks in conjunction with power management. Background Technology

[0002] In scenarios such as warehousing and logistics, manufacturing, and industrial park distribution, construction robots typically perform tasks such as picking up, transporting, and placing goods along different road sections under multi-task concurrent conditions. The road conditions in these scenarios are significantly time-varying. For example, changes in slope, frequent turns, differences in ground friction, speed limits, temporary congestion, and waiting times all affect energy consumption per unit distance. At the same time, load status, acceleration, and braking strategies also change the instantaneous discharge current, thus causing the remaining battery capacity to fluctuate dynamically.

[0003] Existing task allocation schemes typically consider task priority, distance, or estimated completion time, and set remaining battery power trigger conditions to determine whether to allocate a task or switch to charging. Some schemes further introduce power consumption prediction to assist in judging task feasibility and trigger task reassignment based on battery power changes during execution to reduce the risk of mid-journey failure. The above approaches can achieve certain results in static or slowly changing environments, but when road segment parameters and driving information are frequently updated, the predicted power consumption value often needs to be adjusted accordingly. This causes the margin between remaining battery power and predicted power consumption to repeatedly approach or cross the threshold within the critical range, potentially causing the task feasibility judgment to switch between feasible and infeasible. If a fixed margin or single threshold strategy is still used, it is easy to fail to balance between conservative and aggressive approaches under different operating conditions, affecting task completion rate and resource utilization. On the other hand, while triggering task reassignment during the execution period can reduce the probability of single task failure, if there is a lack of quantitative constraints on the frequency of reassignment and closed-loop feedback on energy efficiency costs, multiple task cancellations and re-selection and allocation in a short period of time may induce increased energy consumption that does not directly contribute to effective progress, such as waiting, detours, and repeated start-stops, thereby affecting overall energy efficiency and operational stability.

[0004] Therefore, without denying the effectiveness of existing solutions, it is still necessary to further improve a power-constrained task allocation method for dynamic operating conditions. This method should be able to adaptively form entry and exit thresholds when predictions are updated over time and feasibility fluctuates. It should also combine operational feedback such as the proportion of ineffective energy consumption to control the triggering of task reassignment, so as to achieve a more stable balance between power safety, task continuity and energy efficiency. Summary of the Invention

[0005] To address the shortcomings of existing technologies, the present invention aims to provide a method and system for task allocation in construction robots by combining power management.

[0006] To achieve the above objectives, the present invention provides the following technical solution: A method for task allocation in construction robots by combining power management includes the following steps: Step 1: Obtain parameter information of the driving route, obtain driving information of the construction robot on the driving route, and obtain the remaining battery power of the construction robot; Step 2: Obtain the set of tasks to be executed; input the parameter information and driving information into the power consumption prediction model to obtain the predicted power consumption of each task in the set of tasks to be executed; determine the preset tasks from the set of tasks to be executed. Step 3: Within a preset time window, for each task in the task set to be executed, obtain the updated predicted power consumption, acquire the power feasibility judgment result sequence and the corresponding update timestamp sequence; determine the basic flipping frequency and power feasibility jitter index based on the power feasibility judgment result sequence and the corresponding update timestamp sequence, and determine the equivalent flipping frequency using the basic flipping frequency and the power feasibility jitter index; determine the first power margin and the second power margin according to the equivalent flipping frequency, and determine the entry threshold and exit threshold based on the updated predicted power consumption, the first power margin and the second power margin; acquire the invalid energy consumption ratio, and form an invalid energy consumption ratio sequence according to the preset time window; determine the token replenishment rate and token upper limit of the task reassignment token bucket based on the invalid energy consumption ratio sequence; Step four involves task allocation and dynamic control based on entry and exit thresholds. Dynamic control includes: monitoring the remaining battery power during the execution of the assigned task; when the remaining battery power is less than the exit threshold of the assigned task, determining whether the number of available tokens in the task reassignment token bucket is not less than the preset token consumption amount; if the determination result is yes, executing task reassignment, which includes canceling the assigned task and deducting the corresponding tokens, returning to step three to re-determine the entry and exit thresholds before executing this step for re-screening and allocation; if the determination result is no, issuing a charging task.

[0007] Furthermore, obtaining the updated predicted power consumption includes: when the parameter information or driving information is updated, inputting the updated parameter information or driving information into the power consumption prediction model, updating the predicted power consumption of each task to be executed in the set of tasks to be executed, and obtaining the updated predicted power consumption of each task to be executed in the set of tasks to be executed.

[0008] Furthermore, determining the preset task from the set of tasks to be executed includes: obtaining the task priority of each task to be executed in the set of tasks to be executed; determining the task to be executed with the highest task priority as the preset task; when there are multiple tasks to be executed with the same task priority, determining the task to be executed with the lowest predicted power consumption as the preset task.

[0009] Furthermore, the determination of the basic frequency of the flip includes: determining the flip events based on the sequence of power feasibility judgment results and the corresponding update timestamp sequence; weighting each flip event according to a preset decay time constant to obtain a weighted flip number; and calculating the basic frequency of the flip based on the weighted flip number and the preset time window length.

[0010] Furthermore, the determination of the power feasibility jitter index includes: calculating the flip interval sequence based on the occurrence time of the flip event, and calculating the flip interval feature based on the flip interval sequence; calculating the power difference based on the remaining power and the updated predicted power consumption; calculating the critical approximation feature based on the power difference; calculating the rate of change feature based on the change in the power difference between adjacent update timestamps and the corresponding time interval; and weighting and fusing the flip interval feature, critical approximation feature, and rate of change feature according to preset weights to obtain the power feasibility jitter index.

[0011] Furthermore, the power difference is the difference between the remaining power and the updated predicted power consumption. The power feasibility judgment result includes feasible and infeasible. Within a preset time window, for each task to be executed in the set of tasks to be executed, the power difference is calculated each time the predicted power consumption is updated to the updated predicted power consumption. The power feasibility judgment result is determined based on the power difference, the corresponding update timestamp is recorded, and a sequence of power feasibility judgment results and a corresponding update timestamp sequence are formed in chronological order.

[0012] Furthermore, a flip event is an event in which the power feasibility judgment result switches between two adjacent update timestamps, and the flip interval is the time interval between the occurrence of two adjacent flip events.

[0013] Furthermore, the entry threshold is the sum of the updated predicted power consumption and the first power margin, and the exit threshold is the sum of the updated predicted power consumption and the second power margin. The first power margin is greater than the second power margin, and the difference between the first power margin and the second power margin increases with the increase of the equivalent flip frequency.

[0014] Furthermore, in step three, the ineffective energy consumption ratio sequence is weighted and smoothed to obtain an estimated value, and the rate of change of adjacent preset time windows is calculated. The estimated value and the rate of change are weighted to obtain the reassignment suppression parameter. The attenuation factor is determined by the reassignment suppression parameter, and the token replenishment rate and token limit of the task reassignment token bucket are determined accordingly.

[0015] Furthermore, the system for allocating construction robot tasks in conjunction with power management includes: an information acquisition module, used to acquire parameter information, driving information and remaining power of the driving route; The task prediction module is used to obtain the predicted power consumption of the set of tasks to be executed through the power consumption prediction model and to determine the preset tasks. The threshold token module is used to determine the entry threshold, exit threshold, token replenishment rate, and token limit; The allocation and control module is used to enter the threshold allocation task and exit the task. If the number of threshold tokens is sufficient, the task will be reassigned; otherwise, a charging task will be issued.

[0016] Compared with the prior art, the present invention has the following beneficial effects: This invention updates the predicted power consumption of each task to be executed in a rolling manner when road segment parameter information or driving information is updated, and determines the preset tasks under the comprehensive constraints of task priority and energy consumption prediction, so that the task allocation decision is closer to the real-time operating conditions and energy consumption level, reduces the probability of misallocation caused by prediction lag or single index ranking, and improves the timeliness and reliability of allocation. Based on the sequence of power feasibility judgment results and their update timestamp sequence, flip events are extracted to obtain the basic flip frequency and construct the power feasibility jitter index to form an equivalent flip frequency. Then, the first power margin and the second power margin are adaptively determined by the equivalent flip frequency, and then the entry threshold and the exit threshold are determined. The threshold difference increases with the increase of the equivalent flip frequency, thereby automatically expanding the safety bandwidth in the feasibility critical jitter scenario, reducing the allocation instability caused by repeated triggering near the threshold, and improving the entry threshold and exit conservatism when the risk increases, thus enhancing the execution safety margin. By introducing a time window sequence feedback of the proportion of ineffective energy consumption, the token replenishment rate and token limit of the task reassignment token bucket are determined, and the remaining power is continuously monitored during the execution phase: when the remaining power falls below the exit threshold, reassignment is performed based on whether there are enough tokens, and the tokens are returned to the recalculation threshold for re-selection and allocation, or a charging task is directly issued. This couples "power safety triggering" and "reassignment throttling constraint" in a closed loop, suppressing the accumulation of ineffective energy consumption such as waiting, detours and repeated start-stops caused by frequent reassignment in a short period of time. At the same time, when it is necessary to deal with it, it can still reassign in a timely manner to ensure task continuity and operational stability. Attached Figure Description

[0017] Figure 1 This is a flowchart of the method for task allocation of construction robots in conjunction with power management according to the present invention; Figure 2 This is a schematic diagram of the system structure for task allocation of construction robots in conjunction with power management according to the present invention; Figure 3 This is a schematic diagram of the process of determining the equivalent flip frequency based on the power feasibility judgment result sequence, and determining the entry threshold and exit threshold accordingly, as well as the token bucket parameters for task reassignment in this invention. Figure 4 This is a schematic diagram of the dynamic control process of the present invention, which uses an exit threshold and a task reassignment token bucket to reassign or issue charging tasks during task execution. Detailed Implementation

[0018] To facilitate understanding and ensure the implementation of this invention, the following unified definitions are made for the units of power / energy consumption and parameter constraints involved in this specification: In this specification, remaining power, predicted power consumption, power difference, power margin, entry threshold, and exit threshold are preferably expressed using the same unit of measurement, and optionally can be expressed using available energy. (Unit is) or ), or expressed in terms of state of charge. (Unit is) When using When predicting power consumption in terms of energy, it is preferable to use energy as the basis for calculation. Perform the conversion, among which The available energy capacity of the battery is used to avoid misjudgment of thresholds due to inconsistencies in units; the preset time window length is set to... The update time sequence of predicted power consumption within the preset time window is as follows: The corresponding updated predicted power consumption is The remaining battery power at the corresponding time is The power difference is defined as follows: ,when The result of the power feasibility assessment is recorded as feasible. The feasibility assessment result for electricity consumption is recorded as infeasible, and a sequence of feasibility assessment results for electricity consumption is formed in chronological order. and its corresponding update timestamp sequence Among these, the weight parameters involved in the weighted fusion are preferably non-negative and can further satisfy normalization constraints (e.g., the sum of the weights is 1). (Or normalize according to preset rules) to ensure the scale stability of the fusion result; the coefficients and scale parameters involved in the calculation of exponents, normalization, or amplitude limiting are preferably positive values, and to avoid the denominator being zero or the values ​​being unstable, small positive numbers can be introduced. (For example As a protection item; the above parameters can be calibrated and set according to equipment specifications, historical operating data and real-time requirements of the scenario.

[0019] Example 1: Refer to Figures 1 to 4 The method for task allocation of construction robots in conjunction with power management includes the following steps: Step one involves acquiring parameter information for the travel route, the robot's travel information within that route, and the robot's remaining battery power. This forms the foundational data for task allocation and battery management. By acquiring the parameter information for the travel route and the robot's travel information, we can reflect the sources of energy consumption differences across different road sections and operating states. Furthermore, by obtaining the robot's remaining battery power, we can provide a current battery baseline for subsequent assessments of the feasibility of each task's battery life, as well as for determining entry and exit thresholds. This prevents unreasonable task allocation from being triggered when the battery is insufficient.

[0020] In one specific implementation, parameter information of the travel segment is acquired, along with the construction robot's travel information within that segment and its remaining battery power. Specifically, firstly, based on the travel segment the construction robot will traverse, parameter information of that segment is read or measured. This parameter information includes at least the segment length, gradient change, turning radius, road surface type, speed limit constraints, stopping point location, and potential congestion level. Then, as the construction robot enters or travels along the segment, its travel information is continuously collected. This travel information includes at least the current speed, acceleration, steering changes, braking frequency, waiting time, and load status. Simultaneously, the remaining battery power is obtained by comprehensively measuring the battery terminal voltage, current, and temperature, combined with integral correction during the discharge process. For example, in a scenario where a warehouse passage consists of a straight section plus a 90-degree turn with short stops for waiting, the aforementioned parameter information and travel information can jointly reflect the impact of turning and waiting on energy consumption, while the remaining battery power provides a reliable benchmark for subsequent entry and exit threshold determinations.

[0021] Step two: Obtain the set of tasks to be executed; input parameter information and driving information into the power consumption prediction model to obtain the predicted power consumption of each task in the set; determine the preset task from the set of tasks to be executed; map the "task candidate set" with the "predictable energy consumption" and identify a preset task for priority processing. Obtaining the set of tasks to be executed allows the system to uniformly evaluate multiple tasks to be executed at the same time or within the same period; inputting parameter information and driving information into the power consumption prediction model to obtain the predicted power consumption of each task to be executed enables subsequent threshold construction and feasibility judgment based on the predicted power consumption; determining the preset task from the set of tasks to be executed is used to prioritize task allocation when the power conditions are met, thereby balancing the determinism of task arrangement and overall scheduling efficiency.

[0022] In one specific implementation, a set of tasks to be executed is obtained; parameter information and driving information are input into a power consumption prediction model to obtain the predicted power consumption of each task in the set; a preset task is determined from the set; specifically, the following steps can be performed: receiving transport tasks that have been generated but not completed within the same scheduling cycle, and grouping them according to task type, pick-up and place location, target location, load requirements, and time constraints to form a set of tasks to be executed, and recording the task priority of each task; associating the obtained parameter information and driving information with the expected driving segment corresponding to each task, inputting them into the power consumption prediction model, and outputting the predicted power consumption of each task in the set, so that subsequent selection can simultaneously consider the urgency of the task and the difference in energy consumption; determining the preset task from the set, first comparing the task priorities of each task, and determining the task with the highest task priority as the preset task; when there are multiple tasks with the same task priority, further comparing the predicted power consumption of these tasks, and determining the task with the lowest predicted power consumption as the preset task. For example, when there are both emergency outbound tasks and regular replenishment tasks in the warehouse at the same time, the emergency outbound task can be set as a higher task priority and directly designated as the preset task. When there are two pick-up and drop-off paths with the same priority, one of which involves multiple turns and slope changes resulting in a large predicted power consumption, and the other of which is mainly a straight path resulting in a small predicted power consumption, the task with the smallest predicted power consumption is selected as the preset task under the premise of the same task priority, so as to improve the feasibility and energy efficiency under power-limited conditions.

[0023] In one specific implementation, the power consumption prediction model can be implemented using a "mechanism model baseline + data-driven correction" approach.

[0024] (I) Feature Construction: Constructing feature vectors based on road segment parameter information and driving information. ,For example ;in: The length of the road segment; For slope statistics (e.g., average slope / root mean square slope or segmented slope summary). This refers to statistics on the number of turns or the turning radius. The rolling resistance coefficient is obtained by mapping from the road surface type; For congestion indicators (such as the reduction ratio of average speed relative to the speed limit); For load mass; and These are velocity and acceleration statistics (e.g., mean, variance, quantiles, or root mean square). This indicates the estimated waiting time for service to be suspended.

[0025] (II) Mechanism Baseline Prediction: Discretizing the path into The micro-segment, the first Segment length is The slope angle is The average acceleration is The curvature is (or turning radius) ), assuming total mass Then the equivalent resistance of this segment can be taken as ; in An additional loss factor is added for cornering; based on this, the baseline energy consumption (battery-side consumption) of the mechanism is given, for example... ; in To drive efficiency, For recycling efficiency, To wait for power, To assist load power, The estimated travel time (which can be determined by...) (and velocity statistics estimation).

[0026] (III) Data-driven correction and output: Based on the mechanistic baseline, a data-driven correction term is introduced. Compensation is applied to unmodeled factors (battery aging, localized ground friction, temporary detours, differences in control strategies, etc.) to obtain the predicted power consumption. ; in This can be achieved using linear or nonlinear regression; for example, linear regression. ; ; parameter It can be obtained by fitting the actual energy consumption data of historical tasks; the actual energy consumption can be calculated by BMS sampling, for example... (When only the discharge side is counted, the following can be taken) Furthermore, learning models such as tree models or neural networks can be used to directly output the results. To obtain non-linear expressive capabilities.

[0027] (iv) Online update: When road segment parameter information or driving information is updated, the feature vector is recalculated. And update accordingly and This results in an updated predicted power consumption; optionally, after obtaining new samples... Later Incremental learning (such as sliding window refitting or stochastic gradient updates) can be performed to adapt to environmental changes and improve prediction accuracy.

[0028] Experimental scenario example: A construction material handling robot was used, and four types of road sections were set up within the simulated construction site, as shown in Table 1:

[0029] Each type of road segment was repeatedly run 20 times under no-load, 10kg load, and 20kg load conditions. Data such as voltage, current, speed, acceleration, waiting time, and number of braking operations were collected, and the actual power consumption recorded by the BMS was used as the true value.

[0030] The following is an example of a data table:

[0031] As shown in Table 2, the experimental results show that, under construction site conditions including ramps, turns, waiting, and load changes, the average relative prediction error of this method is 3.04%, which is lower than the 12.59% of the fixed distance estimation method. This indicates that the present invention can improve the accuracy of task power consumption assessment by updating the predicted power consumption based on road segment parameters and driving information.

[0032] Step 3: Within a preset time window, for each task in the task set to be executed, obtain the updated predicted power consumption, acquire the power feasibility judgment result sequence and the corresponding update timestamp sequence; determine the basic flipping frequency and power feasibility jitter index based on the power feasibility judgment result sequence and the corresponding update timestamp sequence, and determine the equivalent flipping frequency using the basic flipping frequency and the power feasibility jitter index; determine the first power margin and the second power margin according to the equivalent flipping frequency, and determine the entry threshold and exit threshold based on the updated predicted power consumption, the first power margin and the second power margin; acquire the invalid energy consumption ratio, and form an invalid energy consumption ratio sequence according to the preset time window; determine the token replenishment rate and token upper limit of the task reassignment token bucket based on the invalid energy consumption ratio sequence; Under the conditions of "predicting dynamic changes in energy consumption" and "fluctuations in the feasibility of electricity use", the system adaptively generates entry and exit thresholds and combines the proportion of ineffective energy consumption to impose frequency constraints on task reassignment.

[0033] 1) Obtain updated predicted power consumption within a preset time window, so that the predicted power consumption can be corrected in a timely manner as parameter information or driving information is updated, and avoid the threshold from deviating from the actual energy consumption due to long-term reliance on outdated predicted power consumption.

[0034] 2) Obtain the sequence of power feasibility judgment results and the corresponding update timestamp sequence, which are used to describe the state evolution process of the same task to be executed from "feasible" to "infeasible" or from "infeasible" to "feasible" under multiple update conditions; this process can characterize whether the power judgment is stable.

[0035] 3) Based on the power feasibility judgment result sequence and the corresponding update timestamp sequence, determine the basic frequency of flipping and the power feasibility jitter index, and determine the equivalent flipping frequency with the basic frequency of flipping and the power feasibility jitter index. Integrate the "frequency of flipping" and the "degree of jitter caused by flipping" into the same metric, provide a unified basis for the adaptive adjustment of subsequent margin and threshold, and make the threshold have stronger safety redundancy for unstable situations.

[0036] 4) Determine the first and second power margins based on the equivalent flip frequency, and determine the entry and exit thresholds based on the updated predicted power consumption, the first and second power margins. Map "predicted power consumption" and "stability risk" together to the thresholds: the entry threshold is used for power threshold control before task allocation, and the exit threshold is used for power lower limit control during task execution, thereby enabling hierarchical management of power risk at different stages.

[0037] 5) Obtain the proportion of ineffective energy consumption and form a sequence of ineffective energy consumption proportions according to a preset time window. Use quantifiable indicators to characterize the changes in the proportion of energy consumption that does not directly contribute to effective execution during operation, so that the system can identify whether there is a trend of low efficiency or large fluctuation in power consumption.

[0038] 6) Determine the token replenishment rate and token upper limit of the task reassignment token bucket based on the invalid energy consumption ratio sequence, and feed back the status and trend of "invalid energy consumption ratio" to the task reassignment constraint mechanism, so that the task reassignment token bucket can throttle and control the upper limit of task reassignment behavior, avoid frequent reassignment when energy efficiency is low or fluctuates greatly, which will lead to further accumulation of invalid energy consumption, and at the same time ensure that there are still available token resources to perform task reassignment when necessary.

[0039] In one specific implementation, within a preset time window, a predicted power consumption update trigger condition is established for each task in the set of tasks to be executed. When the parameter information or driving information is updated, the updated parameter information or driving information is immediately input into the power consumption prediction model to update the predicted power consumption of the task to be executed, thereby obtaining the updated predicted power consumption of the task to be executed. The updated predicted power consumption is then associated with the corresponding update timestamp to ensure that the updated predicted power consumption used for subsequent threshold calculations remains synchronized with the parameter information and driving information of the driving segment. Within a preset time window, whenever the predicted power consumption is updated to the updated predicted power consumption, the power difference is calculated based on the remaining power and the updated predicted power consumption. The power feasibility judgment result is determined based on the power difference, which includes feasible and infeasible. When the power difference is non-negative, it is determined to be feasible; when the power difference is negative, it is determined to be infeasible. At the same time, the corresponding update timestamp is recorded. A sequence of power feasibility judgment results and a corresponding update timestamp sequence are formed in chronological order, so that the power feasibility changes of the same task to be executed within the preset time window can be fully depicted. Reversal events are determined based on the sequence of power feasibility assessment results and their corresponding update timestamp sequences. A reversal event is defined as an event where the power feasibility assessment result switches between two adjacent update timestamps. The occurrence times of each reversal event are used as the event time set. A weighted reversal number is obtained by weighting each reversal event according to a preset decay time constant. Reversal events closer to the end of a preset time window have a higher weight, while those closer to the beginning of the preset time window have a lower weight, making the base reversal frequency more sensitive to recent changes in power feasibility. Then, the base reversal frequency is calculated based on the weighted reversal number and the preset time window length, ensuring that the base reversal frequency reflects the frequency with which the power feasibility assessment result of the task to be executed changes from feasible to infeasible or from infeasible to feasible within the preset time window. Based on the sequence of power feasibility assessment results... and its corresponding update timestamp sequence Determine the flip event: when two consecutive timestamp updates satisfy... At that time, it was believed that in A flip event occurs at this point. Let the current preset time window be... Window length is The set of times when the flip events occurred within this window is: The preset decay time constant is Assign a time decay weight to each flip event. This results in a weighted flip event with a higher weight closer to the end of the preset time window and a lower weight closer to the beginning of the preset time window; the weighted flip number is obtained accordingly. And define the fundamental frequency of the flip as ;when season and .in, Used to adjust the sensitivity to recent flip events. The smaller the value, the stronger the emphasis on recent reversal events; therefore, it is preferable to use a smaller value. and To balance stability and responsiveness, the same order of magnitude is used. While determining the basic frequency of the flipping event, the power feasibility jitter index is further determined: the flipping interval sequence is obtained by calculating the flipping interval based on the occurrence time of the flipping event. The flipping interval is the time interval between the occurrence times of two adjacent flipping events. The flipping interval feature is calculated based on the flipping interval sequence. The flipping interval feature is used to characterize whether the flipping events are concentrated or sparsely distributed. On this basis, the power difference is still the core quantity. The critical approach feature is calculated based on the power difference. The critical approach feature is used to characterize the degree of proximity between the remaining power and the updated predicted power consumption and the risk amplification trend when approaching the critical point. At the same time, the rate of change feature is calculated based on the change in the power difference between adjacent update timestamps and the corresponding time interval. The rate of change feature is used to characterize the steepness of the change in the power difference over time, thereby distinguishing between slow approach and rapid drop. Finally, the flipping interval feature, critical approach feature and rate of change feature are weighted and fused according to preset weights to obtain the power feasibility jitter index, so that the power feasibility jitter index can simultaneously reflect the stability of the flipping interval, the proximity of the critical risk, and the drastic change in the power difference. The equivalent flip frequency is determined by the flip base frequency and the power feasibility jitter index. The equivalent flip frequency is used to combine the flip frequency reflected by the flip base frequency and the jitter intensity reflected by the power feasibility jitter index into a single scale, so as to achieve a unified measurement and adaptive tightening of unstable situations in subsequent margin and threshold calculations. Suppose that the times of the flipping events detected within the preset time window are sorted by time as follows: Then when The time-flipping interval sequence is The flip interval feature can be defined as follows: ;when season ,in To prevent a protection term where the denominator is zero, further, the predicted power consumption update time within the preset time window is set to be... The remaining power is The updated predicted power consumption is The difference in battery power is Critical approximation features are used to characterize the risk amplification trend near the critical point, and are preferably defined as follows: ,in The scale parameter is used to characterize the steepness of the change in the electrical charge difference, and is preferably defined as follows: and take or To facilitate the fusion of different features, it is preferable to perform saturation normalization on the features. ,in Let be a scaling constant; then the power feasibility jitter index is defined as... ,in And preferably satisfy For example, take , , The critical approach characteristic directly reflects the failure risk of "power difference approaching zero," and is given priority in weighting; the reversal interval reflects whether the feasibility involves frequent switching, and is given secondary importance; the rate of change mainly increases when the voltage drops rapidly, and using a smaller weight can avoid being overly sensitive to short-term spikes and causing excessive tightening of the threshold. Based on this, the reversal base frequency is... The frequency of flipping and the jitter index are reflected The intensity of the jitter is combined into a single scale, and the equivalent flip frequency is preferably defined as... ,in This is the adjustment coefficient; when When the equivalent flip frequency degenerates to be determined solely by the fundamental flip frequency, the equivalent flip frequency degenerates into being determined solely by the fundamental flip frequency.

[0040] The first and second power margins are determined based on the equivalent flip frequency. Entry and exit thresholds are determined based on the updated predicted power consumption, the first power margin, and the second power margin. The entry threshold is the sum of the updated predicted power consumption and the first power margin, and the exit threshold is the sum of the updated predicted power consumption and the second power margin. The first power margin is greater than the second power margin, and the difference between the first and second power margins increases with the equivalent flip frequency. This automatically expands the safety bandwidth between the entry and exit thresholds when the equivalent flip frequency is high, reducing misallocation and misassignment caused by frequent switching of power feasibility. For example, when a task to be executed is repeatedly adjusted upwards from a lower value due to slope changes and waiting during a preset time window, causing the power feasibility judgment result to switch between feasible and infeasible, the flip base frequency and the power feasibility jitter index increase synchronously, the equivalent flip frequency increases accordingly, and the first and second power margins increase, with the difference further widening. This makes the entry threshold more stringent and the exit threshold more conservative, thus better meeting the safe execution requirements under fluctuating power conditions. Let the equivalent flip frequency be... The preset reference frequency is Then the normalized equivalent flip frequency can be defined as The first and second battery margins are defined as follows: ; in and To ensure and make Follow Increase it as it increases; furthermore, to avoid excessively large margins leading to overly conservative tasks or excessively small margins leading to increased risks, and Individual amplitude limits can be set: ; ;in Represents the amplitude limiting function. These are the lower and upper limits of the first power margin, respectively, to ensure that the initial margin is not lower than the safe reserve power, and not to be overly conservative; This serves as the lower / upper limit of the second battery capacity margin, used to limit the exit margin and prevent premature / late reassignment. These four parameters can be calibrated according to battery capacity, charging strategy, and risk level, and are typically set as follows: The overall range is higher than The interval creates a hysteresis. Let the updated predicted power consumption corresponding to the task to be executed be... The entry threshold and exit threshold are defined as follows: Thus, as the equivalent flip frequency increases (characterizing enhanced feasibility jitter), the hysteresis bandwidth between the entry threshold and the exit threshold increases accordingly, thereby suppressing frequent triggering and unnecessary reassignment of the critical interval.

[0041] Experiment setup example: Thirty tasks were selected for execution. The robot's remaining battery power was intentionally set close to the critical range of the predicted task power consumption. For example, the remaining battery power was 110Wh, and the predicted task power consumption fluctuated between 95 and 112Wh depending on road condition updates. The following schemes were compared: Compare with Option A: Fix a single power threshold; Comparison with Scheme B: Fixed safety margin threshold; This scheme uses entry / exit thresholds based on equivalent flip frequencies.

[0042] Data table example:

[0043] As shown in Table 3, when the feasibility of task power consumption frequently fluctuates within the critical range, the proposed solution can increase the hysteresis bandwidth between the entry threshold and the exit threshold as the equivalent reversal frequency increases, thereby reducing the number of times the task is cancelled after allocation from 18 times in the fixed single threshold solution to 3 times, and eliminating the situation of insufficient power consumption in the middle.

[0044] The ineffective energy consumption ratio is obtained and formed into an ineffective energy consumption ratio sequence according to a preset time window. The ineffective energy consumption ratio can be obtained from the proportion of energy consumption that does not directly contribute to the effective mission within the preset time window, such as the proportion of energy consumption for waiting, detours, and repeated start-stop operations to the total energy consumption. The ineffective energy consumption ratio sequence is constructed by the ineffective energy consumption ratio of consecutive preset time windows. Based on this, the ineffective energy consumption ratio sequence is weighted and smoothed to obtain an estimate, and the rate of change of adjacent preset time windows is calculated. The estimate is used to suppress occasional spikes, and the rate of change is used to reflect the rising or falling trend of the ineffective energy consumption ratio. The estimate and the rate of change are weighted to obtain a reassignment suppression parameter. The reassignment suppression parameter determines the attenuation factor, and the token bucket for mission reassignment is determined accordingly. The token replenishment rate and token limit are adjusted so that when the estimated proportion of invalid energy consumption is high or the rate of change is increasing, the decay factor decreases accordingly, and the token replenishment rate and token limit decrease accordingly. This reduces the chance of task reassignment being triggered when energy efficiency is low or deteriorating, and avoids frequent cancellation of assigned tasks and re-screening and reassignment, which further induces invalid energy consumption. Conversely, when the estimated proportion of invalid energy consumption is low and the rate of change is stable or decreasing, the decay factor increases accordingly, and the token replenishment rate and token limit increase accordingly. This ensures that the task reassignment token bucket has more available resources to cope with the exit threshold triggering situation caused by a rapid decrease in remaining power or a sudden increase in the predicted power consumption after the update. This achieves a balance between reassignment cost-saving and risk management based on the proportion of invalid energy consumption as feedback.

[0045] Preferred time window The internal energy decomposition calculation is based on the Battery Management System (BMS) and task event logs. Let the sampling time within the window be... Sampling interval The battery terminal voltage and current are respectively Then the discharge power is taken as (Only discharge-side energy is counted), and the total energy consumption within the window is defined.

[0046] Furthermore, define the invalid operation instruction quantity. Used for marking "Ineffective energy consumption" within the range, among which If and only if at least one of the following conditions is met: {1} Idle / Waiting: The robot is in an idle or blocked waiting state (e.g., speed). And no loading / unloading / operation actions were performed); {2} No output was generated after reassignment: in to Task identifiers occurred during the period If a task is switched over, and the switched-over task ends in the task log with "Cancel / Reassign / Abort" instead of "Complete", then the interval from the time the canceled task was last assigned (or started) to the time of cancellation is recorded as an invalid interval; {3} Reverse / No Progress: Maintaining the same task identifier Under the premise of the current task objective point Define distance to target If it appears (Distance increases significantly) or And the duration exceeds If there is no progress for an extended period, the corresponding interval is recorded as an invalid interval. Therefore, the window's invalid energy consumption is defined as follows:

[0047] The percentage of ineffective energy consumption is defined as follows: ;in To prevent Protection terms with a denominator of zero; by definition, we can obtain ,and It can be updated as the window scrolls to be used for subsequent exponential smoothing and suppression mapping.

[0048] In one specific implementation, it is preferable to use a preset time window. Calculate and update continuously. Let the first... Within each window, the proportion of ineffective energy consumption, as defined above, is: Then, exponential weighted smoothing is used to obtain the estimated value of the proportion of ineffective energy consumption. The initial value can be taken as follows: Or take a preset constant The rate of change (trend term) is calculated based on the estimates from adjacent windows.

[0049] in This is used to emphasize the increased suppression of relocation when the proportion of ineffective energy consumption rises. To eliminate dimensional differences, a saturated normalization function is preferably introduced. ( (as a scaling constant), and the estimated proportion of ineffective energy consumption, the rate of change, and the equivalent flip frequency. The relocation suppression parameter is obtained by weighted fusion. ; in And preferably satisfy Based on this, a supplementary attenuation factor (which can also be used for upper limit attenuation) is obtained through nonlinear suppression mapping: ; in The preset suppression coefficient is used. The token bucket parameters are displayed in the window. The inner value is taken as a constant and by Confirmed: Token replenishment rate Token limit ;

[0050] in , These are the preset minimum / maximum replenishment rates. These represent the preset minimum and maximum token limits, respectively. The token bucket state variable is denoted as... (Can be a real number or an integer), and updated according to the following rules at each "potential reassignment decision time": Let the two adjacent decision times be... and (All fall in the same window) (inside), then in First, accumulate according to the replenishment rate and apply an upper limit constraint. ; Then, the deduction rules are used to determine whether reassignment is allowed: Let the preset token amount consumed for each reassignment be... ,like This allows for reassignment and deduction of tokens. ;like This reassignment will be prohibited, and the token will remain unchanged. , It can simultaneously trigger the issuance of a charging task or maintain the execution of the current task. If parameter updates at the window boundary result in a new upper limit... Smaller, preferred execution To ensure the state does not exceed the upper limit, the token bucket can be initially set to a full bucket. Or take This clarifies the entire process of status updates, including supplementation, upper limits, initial values, and deductions. Experiment setup example: Simulate 60 minutes of continuous task scheduling on the same floor of a building. Tasks include material handling, tool delivery, and garbage removal. Dynamic scenarios such as passageway congestion, temporary obstacles, and elevator waiting are artificially introduced. Compare: Comparison plan: If the number of passengers falls below the exit threshold, the passenger will be immediately reassigned. This solution: After the threshold for exiting is reached, it is also necessary to determine whether the available tokens in the token bucket meet the preset consumption amount.

[0051] Example of a supplementary data table:

[0052] As shown in Table 4, the experiment shows that when the proportion of invalid energy consumption increases, the method of this application reduces the token replenishment rate and the token limit, thereby reducing unnecessary reassignment. Compared with the scheme without token bucket limit, the number of reassignments within 60 minutes is reduced from 17 to 6, the proportion of invalid energy consumption is reduced from 22.0% to 10.7%, and the number of completed tasks is increased from 41 to 47.

[0053] Step four involves task allocation and dynamic control based on entry and exit thresholds. Dynamic control includes: monitoring the remaining battery power during the execution of the assigned task; when the remaining battery power is less than the exit threshold of the assigned task, determining whether the number of available tokens in the task reassignment token bucket is not less than the preset token consumption amount; if the determination result is yes, executing task reassignment, which includes canceling the assigned task and deducting the corresponding tokens, returning to step three to re-determine the entry and exit thresholds before executing this step for re-screening and allocation; if the determination result is no, issuing a charging task.

[0054] The entry and exit thresholds are implemented as actual task allocation and runtime protection, and dynamic control is used to achieve "executability guarantee + reassignment suppression".

[0055] 1) Task allocation based on entry threshold can use the entry threshold as a power threshold when allocating tasks, so that the assigned tasks have the power basis to meet the updated predicted power consumption and the first power margin, thereby reducing the probability of insufficient power in the middle of the task.

[0056] 2) During task execution, continuously monitor the remaining power and use the exit threshold as a trigger condition to handle power risks in advance: when the remaining power is less than the exit threshold, it means that the power safety boundary measured by the updated predicted power consumption and the second power margin has been reached, and the dynamic control process needs to be initiated.

[0057] 3) Determine whether the number of available tokens in the task reassignment token bucket is not less than the preset token consumption amount. Use the task reassignment token bucket to constrain the availability of task reassignment and prevent reassignment from being triggered infinitely in a short period of time.

[0058] 4) When the judgment result is yes, the task is reassigned and the process returns to step three to re-determine the entry threshold and exit threshold. On the one hand, the reassignment loop is completed by removing the assigned task and deducting the corresponding token. On the other hand, by returning to step three, the threshold can be recalculated based on the latest updated predicted power consumption, equivalent flip frequency, and invalid energy consumption ratio sequence, thereby improving the adaptability and security of subsequent re-screening and allocation.

[0059] 5) When the judgment result is negative, a charging task is issued. When the reassignment resources of the task reassignment token bucket are insufficient, the system prioritizes the restoration of the construction robot's power to avoid the risk of continuing to run when the power continues to drop and the task cannot be effectively reassigned.

[0060] In one specific implementation, the initialization process for task allocation and dynamic control is based on the entry threshold. That is, after obtaining the entry threshold and exit threshold in step three, the remaining power is compared with the entry threshold. Tasks that meet the entry threshold can be identified as assigned tasks and enter the execution state. At the same time, the exit threshold of the assigned task is bound to the number of available tokens in the task reassignment token bucket as the dynamic control constraint condition for this execution cycle. During the execution of the assigned task, the remaining power is monitored by means of a fixed sampling period or by triggering sampling when parameter information or driving information is updated, so that the change of the remaining power can keep pace with the change of the updated predicted power consumption. Dynamic control and branching are implemented. When the remaining power is detected to be less than the exit threshold of the assigned task, it is immediately determined whether the number of available tokens in the task reassignment token bucket is not less than the preset token consumption. If the result is yes, task reassignment is executed. Task reassignment includes canceling the assigned task and deducting the corresponding tokens. Then, the process returns to step three to redetermine the entry and exit thresholds before re-executing this step for re-screening and reassignment. This allows the re-screening and reassignment to converge based on the latest updated predicted power consumption and the latest invalid energy consumption ratio sequence, which determines the token replenishment rate and token upper limit of the task reassignment token bucket. If the result is no, a charging task is issued to avoid continuing to execute the assigned task when task reassignment resources are insufficient and the remaining power has fallen below the exit threshold, which could lead to mid-process failure or a higher invalid energy consumption ratio.

[0061] For example, if a construction robot encounters temporary congestion in a passageway while performing an assigned task, causing a significant increase in the waiting time in the driving information, and consequently causing the remaining battery power to drop faster than expected and triggering the exit threshold when the remaining battery power falls below the assigned task's exit threshold, then if the number of available tokens in the task reassignment token bucket is still not less than the preset token consumption amount, the assigned task is canceled and the corresponding tokens are deducted. The process then returns to step three to redetermine the entry and exit thresholds before re-selecting and reassigning, allowing the construction robot to execute an updated task with lower predicted power consumption and a better match to the current remaining battery power. Conversely, if the number of available tokens in the task reassignment token bucket is insufficient, a charging task is directly issued, allowing the construction robot to prioritize restoring its remaining battery power and reducing the accumulation of ineffective energy consumption caused by repeated cancellation of assigned tasks and re-selection and reassignment.

[0062] In one specific implementation, to ensure logical closure and engineering feasibility, the system can also perform the following boundary handling: (1) When the remaining power is lower than the entry threshold of all tasks in the task set to be executed, the system will not assign tasks, but will directly issue a charging task or put the robot into a charging waiting state. (2) When SOC is used to represent power consumption and the predicted power consumption is expressed in terms of energy, the unit should be converted according to the available battery capacity before the threshold comparison is performed to avoid misjudgment caused by inconsistency in the units. (3) When the exit threshold is triggered and there are not enough tokens, after the charging task is issued, the canceled / incomplete task can be returned to the set of tasks to be executed and its priority or retry policy can be marked so that it can be rescheduled later.

[0063] Experiment setup example: Construct a task pool containing 100 tasks to be executed, with task types as shown in Table 5:

[0064] Comparing the three options: Option A: Assign tasks based on priority and distance; Option B: Allocate power based on predicted power consumption plus fixed power margin; Option C: The method described in this application.

[0065] Example of a supplementary data table:

[0066] As shown in Table 6, compared with the scheme that allocates tasks based solely on task priority and distance, the scheme in this application increases the task completion rate from 82.0% to 95.0%, reduces the number of misallocations from 14 to 3, and reduces the number of interruptions due to insufficient power to 0. This demonstrates that the scheme in this application can improve the reliability of task allocation and power safety under dynamic working conditions on the construction site.

[0067] Example 2: A system for allocating tasks to construction robots in conjunction with power management, including: an information acquisition module for acquiring parameter information, driving information and remaining power of the driving route; The task prediction module is used to obtain the predicted power consumption of the set of tasks to be executed through the power consumption prediction model and to determine the preset tasks. The threshold token module is used to determine the entry threshold, exit threshold, token replenishment rate, and token limit; The allocation and control module is used to enter the threshold allocation task and exit the task. If the number of threshold tokens is sufficient, the task will be reassigned; otherwise, a charging task will be issued.

[0068] The above description is merely a specific embodiment of this application, but the scope of protection of this application is not limited thereto. Any variations or substitutions that can be easily conceived by those skilled in the art within the scope of the technology disclosed in this application should be included within the scope of protection of this application. Therefore, the scope of protection of this application should be determined by the scope of the claims.

Claims

1. A method for task allocation in construction robots by combining power management, characterized in that, Includes the following steps: Step 1: Obtain parameter information of the driving route, obtain driving information of the construction robot on the driving route, and obtain the remaining battery power of the construction robot; Step 2: Obtain the set of tasks to be executed; input the parameter information and driving information into the power consumption prediction model to obtain the predicted power consumption of each task in the set of tasks to be executed; determine the preset tasks from the set of tasks to be executed. Step 3: Within the preset time window, for each task in the set of tasks to be executed, obtain the updated predicted power consumption, and obtain the power feasibility judgment result sequence and the corresponding update timestamp sequence. The basic frequency of the flip and the power feasibility jitter index are determined based on the power feasibility judgment result sequence and the corresponding update timestamp sequence. The equivalent flip frequency is determined based on the basic frequency of the flip and the power feasibility jitter index. The first power margin and the second power margin are determined based on the equivalent flip frequency. The entry threshold and the exit threshold are determined based on the updated predicted power consumption, the first power margin and the second power margin. Obtain the percentage of ineffective energy consumption and form a sequence of ineffective energy consumption percentages according to a preset time window; The token replenishment rate and token limit of the task reassignment token bucket are determined based on the sequence of ineffective energy consumption ratio. Step 4: Task allocation and dynamic control are performed based on entry and exit thresholds. Dynamic control includes monitoring the remaining battery power during the execution of the assigned tasks. When the remaining battery power is less than the exit threshold of the assigned task, determine whether the number of available tokens in the task reassignment token bucket is not less than the preset token consumption amount; if the result is yes, execute task reassignment, which includes canceling the assigned task and deducting the corresponding tokens, and return to step three to redetermine the entry and exit thresholds before executing this step for re-screening and reassignment; if the result is no, issue a charging task.

2. The method for task allocation of construction robots in conjunction with power management according to claim 1, characterized in that, The updated predicted power consumption is obtained by: when the parameter information or driving information is updated, inputting the updated parameter information or driving information into the power consumption prediction model, updating the predicted power consumption of each task to be executed in the set of tasks to be executed, and obtaining the updated predicted power consumption of each task to be executed in the set of tasks to be executed.

3. The method for task allocation of construction robots in conjunction with power management according to claim 1, characterized in that, Determining a preset task from the set of tasks to be executed includes: obtaining the task priority of each task to be executed in the set of tasks to be executed; determining the task to be executed with the highest task priority as the preset task; when there are multiple tasks to be executed with the same task priority, determining the task to be executed with the lowest predicted power consumption as the preset task.

4. The method for task allocation of construction robots in conjunction with power management according to claim 1, characterized in that, The determination of the basic frequency of the flip includes: determining the flip events based on the sequence of power feasibility judgment results and the corresponding update timestamp sequence; weighting each flip event according to the preset decay time constant to obtain the weighted flip number; and calculating the basic frequency of the flip based on the weighted flip number and the preset time window length.

5. The method for task allocation of construction robots in conjunction with power management according to claim 1, characterized in that, The determination of the power feasibility jitter index includes: calculating the flip interval sequence based on the occurrence time of the flip event, and calculating the flip interval feature based on the flip interval sequence; calculating the power difference based on the remaining power and the updated predicted power consumption; calculating the critical approximation feature based on the power difference; calculating the rate of change feature based on the change in the power difference between adjacent update timestamps and the corresponding time interval; and weighting and fusing the flip interval feature, critical approximation feature, and rate of change feature according to preset weights to obtain the power feasibility jitter index.

6. The method for task allocation of construction robots in conjunction with power management according to claim 5, characterized in that, The power difference is the difference between the remaining power and the updated predicted power consumption. The power feasibility judgment result includes feasible and infeasible. Within a preset time window, for each task to be executed in the set of tasks to be executed, the power difference is calculated each time the predicted power consumption is updated to the updated predicted power consumption. The power feasibility judgment result is determined based on the power difference, the corresponding update timestamp is recorded, and a sequence of power feasibility judgment results and the corresponding update timestamp sequence are formed in chronological order.

7. The method for task allocation of construction robots in conjunction with power management according to claim 5, characterized in that, A flip event is an event in which the power feasibility judgment result changes between two adjacent update timestamps. The flip interval is the time interval between the occurrence of two adjacent flip events.

8. The method for task allocation of construction robots in conjunction with power management according to claim 1, characterized in that, The entry threshold is the sum of the updated predicted power consumption and the first power margin, and the exit threshold is the sum of the updated predicted power consumption and the second power margin. The first power margin is greater than the second power margin, and the difference between the first power margin and the second power margin increases with the increase of the equivalent flip frequency.

9. The method for task allocation of construction robots in conjunction with power management according to claim 1, characterized in that, In step three, the ineffective energy consumption ratio sequence is weighted and smoothed to obtain an estimated value, and the rate of change of adjacent preset time windows is calculated. The estimated value and the rate of change are weighted to obtain the reassignment suppression parameter. The attenuation factor is determined by the reassignment suppression parameter, and the token replenishment rate and token limit of the task reassignment token bucket are determined accordingly.

10. A system for assigning tasks to construction robots in conjunction with power management, applied to the method for assigning tasks to construction robots in conjunction with power management as described in any one of claims 1-9, characterized in that, include: The information acquisition module is used to acquire parameter information, driving information, and remaining battery power for the driving route. The task prediction module is used to obtain the predicted power consumption of the set of tasks to be executed through the power consumption prediction model and to determine the preset tasks. The threshold token module is used to determine the entry threshold, exit threshold, token replenishment rate, and token limit; The allocation and control module is used to enter the threshold allocation task and exit the task. If the number of threshold tokens is sufficient, the task will be reassigned; otherwise, a charging task will be issued.

Images