Business process optimization method and system based on artificial intelligence big data algorithm

Abstract

The invention relates to the technical field of artificial intelligence big data algorithms, and discloses a business process optimization method and system based on an artificial intelligence big data algorithm. In turn, external contextual information is obtained and parsed in response to the combined performance to identify external events associated with the particular operational anomalies. On the basis, causal association between the external event and the specific operation exception can be established, and an optimization strategy of the external event is generated according to the causal association. The method can deeply understand the'accident 'situation which is caused by external emergencies and has strong locality and dynamic characteristics, identify deep causes behind qualitative factors, and establish a complex causal relationship between the qualitative factors and quantitative operation index deterioration.

Description

Technical Field

[0001] This invention relates to the field of artificial intelligence and big data algorithms, and more specifically, to a business process optimization method and system based on artificial intelligence and big data algorithms. Background Technology

[0002] In modern enterprise operations, to improve efficiency and cope with increasingly complex market environments, methods and systems utilizing artificial intelligence and big data technologies have been introduced to optimize business processes. These systems analyze vast amounts of operational data to provide intelligent decision support and execution solutions for various business processes, aiming to overcome the limitations of traditional human experience-based judgment in complex and dynamic scenarios. However, in practical applications, when business processes encounter unexpected situations with strong local and dynamic characteristics caused by external emergencies—such as a sudden shortage of specific highly skilled human resources or nonlinear impacts on local infrastructure—existing systems often struggle to deeply understand the underlying causes behind these qualitative factors. They typically only identify the resulting anomalies in quantitative operational indicators but fail to establish the complex causal relationship between these qualitative factors and the deterioration of quantitative indicators. This results in optimization strategies that are merely stopgap measures and may even trigger new problems, leading to inefficiency and chaos throughout the entire business process.

[0003] For example, in a large fast-moving consumer goods (FMCG) distribution center, an advanced business process optimization system was introduced to significantly improve the efficiency and responsiveness of daily operations. The core of this system lies in its ability to deeply analyze massive amounts of historical operational data, including order volume over the past few years, inventory turnover of various goods, equipment operating status within the distribution center, work efficiency of different shifts, and real-time traffic information and historical delivery records from the external transportation network. By modeling this firsthand information, the system can generate an optimal execution plan down to the minute for daily processes such as receiving, storage, picking, packaging, and shipping. In the initial stages of its operation, the system's performance was outstanding.

[0004] However, when multiple "unexpected" factors occur simultaneously, such as a shortage of specific highly skilled human resources caused by external health events and the nonlinear impact of local extreme weather on specific transportation routes, how can existing business process optimization systems accurately identify the underlying causes of these qualitative factors and establish the complex causal relationship between them and the deterioration of quantitative operational indicators, thereby generating effective optimization strategies based on root causes and avoiding resource allocation imbalances and continuous inefficiency and chaos in business processes due to responses to isolated symptoms? This is a major challenge currently facing the field of business process optimization.

[0005] To address the aforementioned issues, existing technologies urgently need improvement. Summary of the Invention

[0006] This invention provides a business process optimization method and system based on artificial intelligence and big data algorithms. It aims to solve the technical problem that existing business process optimization systems, when faced with "unexpected" situations caused by external emergencies and characterized by strong locality and dynamism, struggle to deeply understand the underlying causes behind qualitative factors and establish the complex causal relationship between these qualitative factors and the deterioration of quantitative indicators. As a result, the optimization strategies proposed by the system are merely treating the symptoms and not the root cause, and may even cause new problems, leading to inefficiency and chaos in the entire business process.

[0007] The technical solution of this application is as follows:

[0008] Firstly, this application discloses a business process optimization method based on artificial intelligence big data algorithms, including:

[0009] Acquire business operation data and analyze it to identify combinations of behaviors that reflect specific operational anomalies;

[0010] In response to combined performance, acquire external contextual information and parse the external contextual information to identify external events related to specific operational anomalies;

[0011] Establish causal relationships between external events and specific operational anomalies;

[0012] An optimization strategy for generating external events based on causal relationships.

[0013] This technical solution enables the identification of combined behaviors reflecting specific operational anomalies from massive amounts of business operation data, and further acquisition of external contextual information to analyze external events related to these anomalies. Based on this, the application can establish a causal relationship between external events and operational anomalies, thereby generating optimization strategies targeting the root causes. This effectively solves the problem in existing technologies of failing to deeply understand the complex causal relationship between qualitative factors and the deterioration of quantitative indicators, avoiding superficial optimization solutions and improving the accuracy and effectiveness of business process optimization.

[0014] Furthermore, in responding to combined performance, acquiring external contextual information, and parsing the external contextual information to identify external events related to specific operational anomalies, the specific steps include:

[0015] When delivery vehicles travel on specific transport routes, the wheels are periodically induced to slip slightly, and wheel speed data and vehicle acceleration data are collected.

[0016] Based on the collected wheel speed data, vehicle acceleration data, and braking pulse parameters, the current road surface micro-friction index is inferred.

[0017] The inferred micro-friction index is compared with a preset safety threshold to identify whether there are hidden low-friction risks on a specific transport route.

[0018] When a hidden low-friction risk is identified, the risk information, along with the vehicle's location, time, and external weather reports, is sent to the business process optimization system to identify external events related to operational anomalies.

[0019] This technical solution enables the accurate inference of the road surface micro-friction index by actively inducing minute wheel slippage and collecting relevant data, thereby identifying hidden low-friction risks that are difficult to detect using traditional methods. This proactive detection mechanism can provide early warnings of potential transportation safety hazards, offering more timely and accurate external contextual information for subsequent business process optimization, effectively improving the depth and breadth of risk identification.

[0020] Based on this, when generating optimization strategies for external events according to the established causal relationships, the specific strategies include:

[0021] When a shortage of specific high-skilled human resources is identified, assess the current workload and historical error rate curves of the manual picking team to determine the safe carrying capacity limit of the manual picking team;

[0022] When a hidden low-friction risk is identified on a specific transport route, a safe recommended speed is calculated and pushed to drivers on the affected road section, and the recommended safe following distance of the fleet is adjusted.

[0023] Based on the safe carrying capacity limit of the manual picking team, the recommended safe speed, and the recommended safe distance of the fleet, the orders to be processed are prioritized.

[0024] Based on the priority ranking results, adjust the allocation of picking tasks and the arrangement of delivery tasks.

[0025] Through this technical solution, this application can provide customized optimization strategies for different types of operational anomalies (such as human resource shortages and transportation risks). By assessing the safe carrying capacity limit of the manual picking team and pushing safe suggested vehicle speeds and distances, this application can achieve intelligent priority ranking of orders and adjust picking and delivery tasks accordingly. In this way, in complex and ever-changing environments, it can effectively balance efficiency and safety, ensure the smooth operation of business processes, and avoid resource allocation imbalances caused by a single strategy.

[0026] Furthermore, when a shortage of specific highly skilled human resources is identified, the current workload and historical error rate curves of the manual picking team are assessed to determine the safe carrying capacity limit of the manual picking team, specifically including:

[0027] Feature analysis is performed on picking tasks that need to be diverted to identify the complexity level of the picking task and the type of operational skills required.

[0028] Query the skill profiles and historical task performance data of the human picking team to assess the real-time effective capacity of the human picking team when handling complex tasks.

[0029] The real-time effective carrying capacity is dynamically adjusted against the historical error rate curve to determine the safe carrying capacity limit of the manual picking team when handling complex tasks.

[0030] This technical solution enables a detailed analysis of the complexity of picking tasks, combined with skill profiles and historical performance data of the manual picking team, to dynamically assess its real-time effective capacity in handling complex tasks. This dynamic adjustment mechanism based on task characteristics and team capabilities makes the determination of the safe capacity limit more accurate and flexible, effectively avoiding resource waste or task backlog caused by simple assessments, and improving the scientific nature of human resource scheduling.

[0031] Building upon the above, when a shortage of specific highly skilled human resources is identified, the current workload and historical error rate curves of the manual picking team are assessed to determine the safe carrying capacity limit of the manual picking team, which also includes:

[0032] Real-time monitoring of the physiological state of members of the manual sorting team to obtain physiological indicators;

[0033] A subjective fatigue survey was conducted on members of the manual sorting team to obtain fatigue information.

[0034] Physiological indicators and fatigue information are combined to infer the real-time fatigue level of members of the manual picking team;

[0035] By combining real-time fatigue levels, current task load, and historical error rate curves, the safe carrying capacity limit for members of the manual picking team is adjusted.

[0036] This technical solution enables real-time inference of the fatigue level of manual picking team members by integrating physiological indicators and subjective fatigue surveys. This deep perception of individual physiological and psychological states allows for more humane and precise adjustments to the safety tolerance limit, effectively preventing decreased work efficiency and increased error rates due to excessive fatigue. This, in turn, ensures employee health while maintaining the stability and efficiency of business processes.

[0037] As a technological improvement, when a shortage of specific highly skilled human resources is identified, the current workload and historical error rate curves of the manual picking team are assessed to determine the safe carrying capacity limit of the manual picking team, which also includes:

[0038] In the manual sorting area, environmental sensing devices are deployed to acquire real-time environmental information of the work area, including noise intensity, light level and temperature information.

[0039] By analyzing the collaborative behavior data of the manual picking team during the operation, the smoothness of internal team collaboration can be identified.

[0040] Context perception questionnaires were administered to members of the manual sorting team to obtain their subjective perceptions of the current environment and teamwork.

[0041] By combining environmental information, collaboration fluency, and subjective perception information, we can infer the potential impact of the work performance of members of the manual picking team.

[0042] By combining the potential impact with the current workload and historical error rate curves, the safe carrying capacity limit of the manual picking team can be dynamically adjusted.

[0043] This technical solution enables the comprehensive acquisition of environmental information about the work area, the smoothness of team collaboration, and the subjective perceptions of team members by deploying environmental sensing devices, analyzing team collaboration behavior data, and conducting context-aware questionnaires. This fusion of multi-dimensional information allows for more accurate inferences about the potential impact on the performance of manual picking team members. This, in turn, allows for more precise and dynamic adjustments to the safety tolerance limit, effectively addressing the impact of environmental changes and team collaboration on work efficiency, and further enhancing the adaptability and robustness of the business process.

[0044] In one implementation, environmental sensing devices are deployed in the manual picking area to acquire environmental information of the area in real time, specifically including:

[0045] In the manual picking area, multiple environmental sensing devices are deployed, including noise sensors, light sensors, and temperature sensors. These environmental sensing devices form a distributed network to collect environmental information of the work area in real time.

[0046] The environmental information collected by the distributed network is fused with the micro-environmental information obtained by wearable environmental sensors configured on the manual picking team to construct an environmental information distribution map of the work area.

[0047] Based on the environmental information distribution map, identify areas where environmental information is missing or inaccurate;

[0048] When areas with missing or inaccurate environmental information are identified, the missing or inaccurate environmental information is supplemented or corrected based on the environmental information distribution map.

[0049] This technical solution enables the construction of a detailed environmental information distribution map of the work area by deploying distributed environmental sensing devices and combining them with wearable environmental sensors. This multi-source data fusion and dynamic completion / correction mechanism ensures the comprehensiveness and accuracy of environmental information, effectively solving the problems of blind spots or data bias that may exist with single sensors, and providing reliable environmental data support for subsequent operational performance evaluation.

[0050] Based on the above, multiple environmental sensing devices are deployed in the manual picking area. These devices include noise sensors, light sensors, and temperature sensors. They form a distributed network to collect real-time environmental information about the work area, specifically including:

[0051] In each environmental sensing device in the distributed network, a self-test program is periodically triggered, which generates sensor response data through an internal reference source or a preset standard signal.

[0052] Compare sensor response data with historical calibration data of environmental sensing devices to identify whether there is performance drift or calibration deviation in environmental sensing devices;

[0053] When performance drift or calibration deviation is detected, the data output of the environmental sensing device is adjusted according to the degree of performance drift or calibration deviation in order to perform adaptive calibration.

[0054] The system continuously monitors environmental information collected by environmental sensing devices. When abnormal fluctuations occur in the environmental information that are significantly inconsistent with data from surrounding devices or historical trends, the abnormal detection mechanism is triggered, and the abnormal fluctuations are marked.

[0055] This technical solution enables adaptive calibration of environmental sensing devices through periodic self-checks and comparison with historical calibration data, effectively addressing the performance drift issue that may occur during long-term sensor operation. Simultaneously, the anomaly detection mechanism can promptly identify and flag abnormal fluctuations in environmental information, ensuring the real-time nature and accuracy of environmental data and providing a more reliable data foundation for business process optimization.

[0056] In one implementation, environmental information collected by a distributed network is fused with wearable environmental sensors configured on the manual picking team to construct an environmental information distribution map of the work area, specifically including:

[0057] Extract the first timestamp and first spatial identifier of the environmental information collected by the distributed network, and extract the second timestamp and second spatial identifier of the micro-environment information acquired by the wearable environmental sensor;

[0058] The first timestamp is compared with the second timestamp. When the difference between the compared timestamps exceeds a preset first threshold, the first timestamp and the second timestamp are calibrated.

[0059] The first spatial identifier is compared with the second spatial identifier. When the difference between the spatial identifiers exceeds a preset second threshold, the first spatial identifier and the second spatial identifier are calibrated.

[0060] Based on the calibrated timestamps and spatial identifiers, the environmental information collected by the distributed network and the micro-environment information obtained by the wearable environmental sensors are mapped to a unified grid cell, and the information within each grid cell is aggregated to construct an environmental information distribution map of the work area.

[0061] This technical solution enables the application to ensure the spatiotemporal consistency of multi-source environmental information by calibrating the timestamps and spatial identifiers of data collected from distributed networks and wearable environmental sensors. This refined data fusion and gridded aggregation processing can construct a high-precision, unbiased distribution map of the work area's environmental information, providing more accurate and comprehensive environmental perception data for subsequent business process optimization.

[0062] Secondly, this application also discloses a business process optimization system based on artificial intelligence big data algorithms, including:

[0063] The input end is used to acquire business operation data and analyze the business operation data to identify combinations of behaviors that reflect specific operational anomalies.

[0064] The parsing end is used to respond to the combined performance, obtain external context information, and parse the external context information to identify external events related to specific operational anomalies;

[0065] The generation end is used to establish causal relationships between external events and specific operational anomalies; based on these causal relationships, it generates optimization strategies for external events.

[0066] Through this technical solution, the system of this application can achieve in-depth analysis of business operation data, intelligent parsing of external contextual information, and establishment of causal relationships by working collaboratively between the input end, parsing end, and generation end, ultimately generating effective optimization strategies. This system architecture effectively integrates artificial intelligence and big data algorithms, fundamentally solving the problem that existing systems cannot identify deep-seated causes and generate effective optimization strategies in complex and dynamic scenarios, thus improving the automation and intelligence level of business process optimization.

[0067] Beneficial effects

[0068] This application discloses a business process optimization method and system based on artificial intelligence big data algorithms. It acquires business operation data and identifies combinations of behaviors reflecting specific operational anomalies. Then, it responds to these combinations of behaviors to acquire and analyze external contextual information to identify external events related to the specific operational anomalies. Based on this, the application can establish a causal relationship between external events and specific operational anomalies, and generate optimization strategies for external events according to this causal relationship. This method can deeply understand "unexpected" situations triggered by external emergencies, characterized by strong locality and dynamism, identify the underlying causes behind qualitative factors, and establish complex causal relationships between these qualitative factors and the deterioration of quantitative operational indicators. Compared to existing technologies, this application can provide effective optimization strategies based on root causes, avoiding the problems of resource allocation imbalance and continuous inefficiency and chaos in business processes caused by reacting to isolated symptoms. This significantly improves the accuracy, effectiveness, and robustness of business process optimization, and solves the limitations of traditional human experience-based judgment in complex and dynamic scenarios. Attached Figure Description

[0069] Figure 1 This is a flowchart of a business process optimization method based on artificial intelligence big data algorithms provided in an embodiment of the present invention;

[0070] Figure 2 This is a flowchart of a method for identifying external events related to specific operational anomalies provided by an embodiment of the present invention;

[0071] Figure 3 This is a schematic diagram of the structure of a business process optimization system based on artificial intelligence big data algorithms provided in an embodiment of the present invention. Detailed Implementation

[0072] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0073] Reference Figure 1 , Figure 1 This is a flowchart of a business process optimization method based on artificial intelligence big data algorithms provided in an embodiment of the present invention, including:

[0074] S11, acquire business operation data and analyze the business operation data to identify combined behaviors that reflect specific operational anomalies;

[0075] S12, in response to the combined performance, acquire external context information and parse the external context information to identify external events related to specific operational anomalies;

[0076] S13, Establish the causal relationship between external events and specific operational anomalies;

[0077] S14, Based on causal relationships, generate an optimization strategy for external events.

[0078] This application, by introducing the acquisition and analysis of external contextual information, can identify external events related to specific operational anomalies and further establish causal relationships between these external events and the specific operational anomalies. This generates more targeted and fundamental optimization strategies, effectively avoiding the problem of traditional systems that only address the symptoms and not the root cause, and improving the resilience and adaptability of business processes.

[0079] The business process optimization method proposed in this application aims to deeply perceive, analyze, and make decisions about complex business operation environments through artificial intelligence and big data technologies.

[0080] "Business operation data" refers to various structured and unstructured data generated by an enterprise during its daily operations, such as order volume, inventory levels, logistics tracking, employee performance, and equipment status. This data forms the basis for assessing the health of the business. "Combined manifestations of specific operational anomalies" refers to the simultaneous occurrence of multiple business indicators or events in a specific pattern, indicating potential business problems, such as order backlog and delivery delays occurring simultaneously. "External contextual information" refers to external environmental factors affecting business operations, such as weather, traffic, and market dynamics. "External events" refer to specific events that have a real impact on business operations, as analyzed from external contextual information. "Causal correlation" refers to the logical relationship between external events and specific operational anomalies, i.e., the external event is the cause of the specific operational anomaly. "Optimization strategies" refer to a series of action plans developed to restore or improve business efficiency in response to external events and the resulting operational anomalies.

[0081] The core of the business process optimization method in this application lies in the in-depth mining and utilization of business operation data, external contextual information, and the causal relationships between them.

[0082] First, various methods can be employed to acquire and analyze business operation data to identify combinations of behaviors reflecting specific operational anomalies. For example, internal data sources such as Enterprise Resource Planning (ERP), Customer Relationship Management (CRM), and Supply Chain Management (SCM) systems can be used to collect order data, inventory data, production data, sales data, and logistics data in real time or periodically. This data can be stored in a data warehouse or data lake. Subsequently, statistical analysis methods, such as mean, variance, and correlation analysis, or more complex machine learning algorithms, such as anomaly detection models (e.g., based on Isolated Forest, Local Outlier Factor (LOF), or Gaussian Mixture Model (GMM), can be used to process this data. By setting warning thresholds or pattern matching rules, when multiple business indicators simultaneously deviate from the normal range, forming a specific combination pattern, combinations of behaviors reflecting specific operational anomalies can be identified. For example, when order completion rate continues to decline, accompanied by an increase in customer complaint rate and abnormal fluctuations in inventory turnover rate, the system can identify this as a specific combination of operational anomalies.

[0083] Secondly, in responding to combined performance, acquiring external contextual information, and parsing this information to identify external events related to specific operational anomalies, the following methods can be adopted. For example, web crawling technology can be used to obtain real-time weather forecasts, traffic conditions, breaking news, and other information from public websites (such as meteorological bureau websites, transportation department websites, and news media websites). Simultaneously, data such as market reports, industry analyses, and policy updates provided by third-party data service providers can also be obtained. This external contextual information can take various forms, including text, images, and videos. To parse this heterogeneous data, Natural Language Processing (NLP) technology can be used to perform entity recognition, sentiment analysis, and event extraction on textual information, such as identifying keywords like "heavy rain" and "road closure" and their associated events. For image and video information, computer vision technology can be used for scene recognition and event detection. By comparing and associating the parsed external contextual information with the identified specific operational anomalies, external events related to the operational anomalies can be identified. For example, when an operational anomaly combination of "delivery delays" is identified, the system will check for external events such as "heavy rain" or "road closures" and identify them as potential related external events.

[0084] Furthermore, various modeling methods can be employed to establish causal relationships between external events and specific operational anomalies. For example, an expert knowledge base can be used to predefine a series of known causal rules, such as "extreme weather leads to logistics delays." When a corresponding external event and operational anomaly are identified, the system can directly match these rules. Further, machine learning-based causal inference models can be used, such as Granger causality tests, structural causality models (SCM), or Bayesian networks. These models can automatically learn and establish causal relationships by analyzing the temporal relationships and statistical dependencies between external events and operational anomalies in historical data. For instance, by analyzing the probability and duration of "delivery delays" after each "road closure" event in the past year, the system can quantify the causal impact of "road closures" on "delivery delays."

[0085] Finally, in generating optimization strategies for external events based on causal relationships, decision support systems or reinforcement learning models can be utilized. For example, when a causal relationship is identified where "road closure" leads to "delivery delay," the system can generate optimization strategies such as "adjusting delivery routes," "notifying customers of delays," and "adding backup vehicles" based on a pre-set strategy library. If reinforcement learning is used, the system can continuously learn and optimize the strategy generation process by simulating the execution effects of different optimization strategies in different scenarios to find the optimal solution. For example, when faced with an external event of "shortage of highly skilled human resources," the system can generate strategies such as "reassigning tasks," "initiating an emergency recruitment process," and "training existing employees" based on causal relationships, and evaluate the impact of these strategies on business efficiency and costs to select the best strategy.

[0086] The business process optimization method of this application acquires business operation data and identifies a combination of specific operational anomalies. In response to the combination of performance, it acquires and parses external contextual information to identify external events related to specific operational anomalies. Then, it establishes a causal relationship between external events and specific operational anomalies, and finally generates optimization strategies for external events based on the causal relationship, thus forming a closed-loop intelligent optimization process.

[0087] The business process optimization method presented in this application represents a significant advancement compared to existing technologies. Traditional systems often only identify quantitative operational metric anomalies, but struggle to deeply understand the underlying causes, especially when faced with qualitative factors triggered by external unforeseen events. For example, existing systems might only detect "delivery delays," but fail to effectively identify the root cause as "road closures due to extreme weather." This application, by introducing the acquisition and analysis of external contextual information, can identify external events related to specific operational anomalies and further establish causal relationships between these events and the specific operational anomalies. This in-depth analysis of causal relationships allows the system to fundamentally understand the crux of the problem, thereby generating more targeted and fundamental optimization strategies. For instance, when "a shortage of specific high-skilled human resources" is identified as the root cause of "increased picking error rates," the system will not merely suggest increasing the number of picking personnel, but will further consider factors such as personnel skill matching and task complexity assessment, thus proposing a more refined solution. This effectively avoids the problem of traditional systems only addressing the symptoms and not the root cause, improving the resilience and adaptability of business processes.

[0088] In some embodiments described above, this application proposes acquiring and parsing external contextual information in response to combined performance to identify external events related to specific operational anomalies. However, in actual business operations, some external contextual information may be obscure and difficult to acquire and identify directly through conventional means, such as subtle changes in road surface friction. This may lead to untimely or inaccurate identification of operational anomalies. If these problems are not addressed, the business process optimization system may be unable to provide early warnings and interventions for potential risks, thereby affecting overall operational efficiency and safety.

[0089] In this regard, refer to Figure 2 , Figure 2 This is a flowchart of a method for identifying external events related to specific operational anomalies provided in an embodiment of the present invention. S12 includes:

[0090] S121, when the delivery vehicle is traveling on a specific transport route, periodically induces slight slippage of the wheels and collects wheel speed data and vehicle acceleration data.

[0091] S122, Based on the collected wheel speed data, vehicle acceleration data, and braking pulse parameters, infer the current road surface micro-friction index;

[0092] S123, compare the inferred micro-friction index with a preset safety threshold to identify whether there is a hidden low-friction risk in the specific transport trunk line;

[0093] S124, when the hidden low-friction risk is identified, the risk information, along with the vehicle location, time, and external weather report, is sent to the business process optimization system to identify external events related to the operational anomaly.

[0094] Specifically, when delivery vehicles travel along specific transport routes, periodically inducing minute wheel slippage refers to briefly creating extremely small relative movements between the wheels and the road surface through the vehicle's braking or power system, without affecting the vehicle's normal driving safety. The purpose is to detect the actual friction characteristics between the road surface and the tires through this controlled minute slippage. Collecting wheel speed data and vehicle acceleration data involves using wheel speed sensors and acceleration sensors installed on the vehicle to obtain real-time information on wheel rotation speed and changes in the overall vehicle motion state. This data forms the basis for subsequent inferences about the road surface friction index.

[0095] Based on collected wheel speed data, vehicle acceleration data, and braking pulse parameters, the micro-friction index of the current road surface is inferred. This can be understood as using a vehicle dynamics model and friction calculation algorithm, combined with real-time collected wheel speed data, acceleration data, and braking pulse parameters emitted by the vehicle's braking system (such as braking pressure and braking time), to accurately calculate the micro-friction coefficient at the current contact point between the road surface and the tire. This micro-friction index can quantify the slipperiness or adhesion of the road surface.

[0096] In practical applications, comparing the inferred micro-friction index with a preset safety threshold to identify hidden low-friction risks on specific transport routes involves comparing the real-time calculated micro-friction index with a pre-set minimum friction threshold representing safe driving conditions. If the inferred friction index is lower than this threshold, it indicates that there is a potential low-friction risk on that road section that is difficult to detect with the naked eye, such as thin ice, slippery road surfaces, or oil stains.

[0097] When a hidden low-friction risk is identified, the risk information, along with the vehicle's location, time, and external weather report, is sent to the business process optimization system to identify external events related to the operational anomaly. This means that once a low-friction risk is confirmed, the system immediately sends detailed information about the risk (including risk level, specific location, and time of occurrence) and supporting information such as the prevailing weather conditions to the central business process optimization system via the vehicle communication module. The business process optimization system treats this information as an external event and associates it with potential operational anomalies (such as traffic accidents, delays, etc.).

[0098] This application's solution actively induces minute wheel slippage and combines multi-source data (wheel speed data, acceleration data, and braking pulse parameters) to infer the micro-friction index of the road surface, thus overcoming the limitations of traditional methods in accurately identifying hidden low-friction risks in real time. Through this technical solution, this application can achieve early and accurate identification of potential, hidden external risk events in business operations. Compared to relying solely on passively acquired external contextual information, this solution, through an active detection mechanism, significantly improves the ability and timeliness of identifying subtle external events such as low-friction risks on the road surface. This allows the business process optimization system to intervene earlier, providing early warnings and interventions for operational anomalies that may be caused by these hidden risks, thereby effectively reducing accident rates, minimizing transportation delays, ensuring delivery safety, and improving the resilience and efficiency of the overall business process.

[0099] In some of the embodiments described above in this application, optimization strategies for generating external events based on causal relationships are proposed. However, in the actual business process optimization process, generating general strategies based solely on causal relationships may not be sufficient to cope with complex and ever-changing abnormal operational situations. In particular, when multiple external events occur simultaneously or require refined intervention, the lack of specific and operable strategy generation mechanisms may lead to poor optimization results or delayed response.

[0100] In response, this application further proposes an optimization strategy for generating external events based on the established causal relationships, including:

[0101] When a shortage of specific high-skilled human resources is identified, assess the current workload and historical error rate curves of the manual picking team to determine the safe carrying capacity limit of the manual picking team;

[0102] When a hidden low-friction risk is identified on a specific transport route, a safe recommended speed is calculated and pushed to drivers on the affected road section, and the recommended safe following distance of the fleet is adjusted.

[0103] Based on the safe carrying capacity limit of the manual picking team, the recommended safe speed, and the recommended safe distance of the fleet, the orders to be processed are prioritized.

[0104] Based on the priority ranking results, adjust the allocation of picking tasks and the arrangement of delivery tasks.

[0105] Specifically, when a shortage of specific highly skilled human resources is identified, it is necessary to assess the current workload and historical error rate curve of the manual picking team to determine the team's safe carrying capacity limit. The current workload refers to the total amount and complexity of picking tasks being processed or about to be processed by the manual picking team within a specific time period. The historical error rate curve reflects the team's accuracy and efficiency in completing tasks at different workload levels, typically showing a trend of increasing error rate as the workload increases. By comprehensively analyzing these two data points, the maximum workload that the manual picking team can withstand while maintaining established service quality and efficiency standards—that is, the safe carrying capacity limit—can be dynamically inferred. For example, machine learning models can be used, combining historical task data, personnel skill profiles, and error records, to predict the team's carrying capacity under current staffing and task types.

[0106] Furthermore, when a hidden low-friction risk is identified on a specific transport route, it is necessary to calculate and push recommended safe speeds to drivers on the affected road sections, and adjust the recommended safe following distances within the convoy. Hidden low-friction risks refer to conditions where the road surface friction coefficient is lower than normal but difficult to detect with the naked eye, such as thin ice, slippery surfaces, or oil stains. For such risks, the system will accurately calculate the recommended safe driving speed for vehicles under these low-friction conditions based on real-time road conditions, weather data, and vehicle sensor data, and push this information to relevant drivers via the in-vehicle system or mobile application. Simultaneously, to ensure the overall driving safety of the convoy, the recommended safe following distances between vehicles within the convoy will be adjusted accordingly to allow for longer reaction time and braking distance.

[0107] Based on this, the proposed solution prioritizes orders to be processed according to the safe carrying capacity limit of the manual picking team, the recommended safe speed, and the recommended safe distance of the fleet. This means that the system will comprehensively consider the availability of internal resources (human resources) and the risks of the external environment (road conditions) to intelligently evaluate all pending orders. For example, for orders that require highly skilled picking and whose destinations are located in high-risk areas, their priority may be adjusted, or they may be assigned to pickers with higher skills and lower workloads, and a safer delivery route may be planned.

[0108] Finally, based on the priority ranking, the allocation of picking tasks and the scheduling of delivery tasks are adjusted. This includes assigning higher-priority picking tasks to pickers with lower current workloads and matching skills to ensure timely and efficient completion. Simultaneously, delivery task scheduling is optimized based on road risks and recommended vehicle speeds. For example, safer alternative routes are selected for orders on high-risk road sections, or delivery times are adjusted to avoid inclement weather, ensuring safe and timely delivery of goods.

[0109] This application's solution, through detailed analysis and quantification of specific operational anomalies, combines human resource conditions with external environmental risks to construct a multi-dimensional optimization decision-making model. Through this technical solution, this application provides a more refined, context-aware method for business process optimization. This method not only identifies operational anomalies and establishes causal relationships, but more importantly, it can generate highly actionable and targeted optimization strategies for specific external events. This enables business processes to respond quickly to fluctuations in human resources or external environmental risks, effectively balancing efficiency and safety, reducing operating costs, improving customer satisfaction, and enhancing the resilience and adaptability of the entire business system.

[0110] In some embodiments described above, this application proposes assessing the current workload and historical error rate curves of a manual picking team when a shortage of specific high-skilled human resources is identified, in order to determine the safe carrying capacity limit of the manual picking team. However, in its implementation, relying solely on macroscopic workload and historical error rates may fail to accurately capture the complexity differences of different picking tasks and the skill strengths of team members, leading to biases in the assessment of the team's actual carrying capacity and affecting the effectiveness of subsequent optimization strategies. Therefore, this application further proposes a more refined assessment method, aiming to dynamically adjust the safe carrying capacity limit through in-depth analysis of task characteristics and team skill profiles, thereby achieving more accurate resource allocation.

[0111] When a shortage of specific high-skilled human resources is identified, the current workload and historical error rate curves of the manual picking team are assessed to determine the safe carrying capacity limit of the manual picking team, including:

[0112] Feature analysis is performed on picking tasks that need to be diverted to identify the complexity level and the type of operational skills required for the picking tasks;

[0113] Query the skill profile and historical task performance data of the manual picking team to assess the real-time effective carrying capacity of the manual picking team when handling complex tasks;

[0114] The real-time effective carrying capacity and historical error rate curve are dynamically adjusted to determine the safe carrying capacity limit of the manual picking team when handling the complex task.

[0115] Specifically, feature analysis of picking tasks requiring decentralization involves a detailed breakdown and evaluation of each task to be assigned to the human picking team. This analysis aims to identify the inherent complexity level of each picking task, such as the number and types of goods involved, the length of the picking path, the volume and weight of the goods, and special operational requirements (such as handling fragile items or temperature control requirements). Simultaneously, it is necessary to identify the specific operational skills required to complete these tasks, such as fine motor skills, heavy-duty handling skills, and specific equipment operation skills. The purpose is to provide foundational data for subsequent team capability assessment and task matching.

[0116] The process of querying the skill profiles and historical task performance data of the manual picking team to assess their real-time effective capacity in handling complex tasks can be understood as the system accessing a pre-established database of team member skills and historical work records. The skill profiles may include information such as each team member's professional skills, training certifications, and experience level. Historical task performance data records indicators such as efficiency, accuracy, and error types of team members when handling tasks of varying complexity in the past. By comprehensively analyzing this data, the system can quantitatively assess the team's current actual processing capacity and efficiency when facing tasks of specific complexity and skill requirements. The aim is to provide a reference for the team's actual operational capabilities in different scenarios.

[0117] In practical applications, dynamically adjusting the real-time effective capacity against the historical error rate curve to determine the safe upper limit of the manual picking team's workload when handling complex tasks involves combining the team's real-time effective capacity with its historical error rate curve. The historical error rate curve reflects the error trend of the team under different loads or task complexities. Through dynamic adjustment, the maximum safe load the team can achieve under a specific task scenario can be predicted based on the current real-time effective capacity—that is, the amount of work the team can handle without a significant increase in the error rate. For example, if the real-time effective capacity is high but the historical error rate curve shows a sharp increase in the error rate under a specific high load, the safe upper limit will be appropriately lowered to avoid potential operational risks. The aim is to ensure that operational quality and safety are maintained or improved while optimizing business processes.

[0118] This application's solution, by introducing feature analysis of picking tasks, can meticulously identify the complexity levels and required operational skill types of different tasks, thus providing a precise task requirement profile for subsequent team capability assessment. Through the above technical solution, this application can achieve a more accurate and dynamic assessment of the safe carrying capacity limit of a manual picking team. Specifically, by performing feature analysis on picking tasks, assessment biases caused by differences in task complexity can be avoided; by querying the team's skill profile and historical task performance data, the true capabilities of the team in handling specific tasks can be more accurately reflected; and by dynamically correcting the real-time effective carrying capacity against historical error rate curves, it further ensures that the determined safe carrying capacity limit not only considers efficiency but also operational quality and safety. Therefore, when a shortage of highly skilled human resources is identified, a more reasonable, efficient, and safe business process optimization strategy can be generated based on a more reliable carrying capacity limit, effectively improving the accuracy of resource allocation and the stability of overall operations, and reducing error rates and operational risks caused by human resource shortages or improper task allocation.

[0119] In some embodiments described above, when a shortage of specific highly skilled human resources is identified, the safe carrying capacity limit of the manual picking team is determined by evaluating its current workload and historical error rate curves. However, in practice, relying solely on workload and historical error rates may not fully reflect the real-time working status of team members, particularly the impact of physiological and psychological fatigue levels on work efficiency and error rates. This limitation may lead to misjudgment of the team's actual carrying capacity, thereby affecting the accuracy and effectiveness of business process optimization. Therefore, this application further proposes a more refined evaluation method that dynamically adjusts the safe carrying capacity limit of the manual picking team by real-time monitoring of team members' physiological state and subjective fatigue.

[0120] When a shortage of specific high-skilled human resources is identified, the current workload and historical error rate curves of the manual picking team are assessed to determine the safe carrying capacity limit of the manual picking team, including:

[0121] The physiological state of the members of the manual sorting team is monitored in real time to obtain physiological indicators;

[0122] A subjective fatigue survey was conducted on the members of the manual sorting team to obtain fatigue information;

[0123] The physiological indicators and fatigue information are fused together to infer the real-time fatigue level of the members of the manual sorting team;

[0124] By combining the real-time fatigue level, the current task load, and the historical error rate curve, the safe carrying capacity limit of the members of the manual picking team is adjusted.

[0125] Specifically, the physiological state of members of the manual sorting team is monitored in real time to obtain physiological indicators that reflect their physical condition and fatigue level. These physiological indicators may include, but are not limited to, heart rate, heart rate variability, skin conductance, body temperature, eye movement data (such as blinking frequency and pupil size changes), and electroencephalogram (EEG) activity. This data can be continuously and imperceptibly collected by wearing smart wearable devices, such as smart bracelets, smartwatches, smart glasses, or clothing with integrated sensors. Heart rate variability effectively reflects the activity state of the autonomic nervous system and is closely related to an individual's mental stress and fatigue level; changes in skin conductance may indicate emotional excitement or stress response; and eye movement data directly reflects visual fatigue and the degree of concentration.

[0126] Simultaneously, a subjective fatigue survey was conducted on members of the manual sorting team to obtain their subjective perceptions of their own fatigue levels. This survey could be conducted by periodically sending short questionnaires to team members or through a voice interaction system. For example, the Visual Analogue Scale (VAS) or Likert Scale could be used to quantify their fatigue feelings, asking questions such as "How fatigued are you currently feeling?" and "How well do you concentrate?" This subjective information is an important supplement to objective physiological indicators, because an individual's perception of fatigue can be influenced by various factors and is not always perfectly synchronized with physiological indicators.

[0127] Furthermore, the acquired physiological indicators and fatigue information are fused to infer the real-time fatigue level of members of the manual selection team. The fusion process can employ various data fusion algorithms, such as machine learning-based models (e.g., support vector machines, neural networks), Bayesian networks, or fuzzy logic systems. These models can learn the complex relationship between physiological indicators and subjective fatigue information, and comprehensively judge an individual's fatigue state. For example, when heart rate variability decreases, blink rate increases, and subjectively reported fatigue is high, the system can infer that the member is in a state of moderate or severe fatigue.

[0128] Finally, by combining the inferred real-time fatigue level, current workload, and historical error rate curves, the safe carrying capacity limit for members of the manual picking team is adjusted. This means that real-time human condition parameters are introduced into the traditional assessment based on workload and historical performance. For example, when a member is inferred to be at a high fatigue level, their safe carrying capacity limit may be lowered even if their current workload is not high, to avoid an increase in the error rate due to fatigue. Conversely, if a member is in good condition, their carrying capacity limit may be maintained or appropriately increased. The historical error rate curve provides a benchmark for long-term trends and individual differences, making the adjustment process more accurate and personalized.

[0129] This application's solution, by introducing real-time monitoring of the physiological state and subjective fatigue survey of manual picking team members, can obtain more comprehensive and dynamic individual fatigue information. Through this technical solution, this application can achieve a more accurate and dynamic assessment of the safe carrying capacity limit of the manual picking team. Compared to methods relying solely on static task load and historical data, this solution incorporates real-time considerations of human physiological and psychological states, significantly improving the real-time nature and accuracy of the assessment. This effectively avoids decreased operational efficiency and increased error rates due to team member fatigue, thereby optimizing the rationality of task allocation and reducing operational risks. This dynamic correction mechanism based on real-time fatigue levels not only ensures the health and safety of team members but also maximizes the overall team effectiveness while ensuring operational quality, enhancing the overall resilience and adaptability of the business process.

[0130] Traditional business process optimization methods typically rely solely on the team's current workload and historical error rate curves when assessing the safe carrying capacity of manual picking teams. However, this approach may fail to adequately consider the dynamic changes in the work environment, the efficiency of collaboration among team members, and the impact of individual subjective perceptions on actual performance. Without taking these crucial factors into account, the determined safe carrying capacity may deviate from the team's true capabilities, leading to unreasonable task allocation and potentially increasing error rates and employee fatigue. To address this, this application proposes a more comprehensive and dynamic assessment method that integrates multi-dimensional information to accurately adjust the safe carrying capacity of manual picking teams.

[0131] In some embodiments of this application, the step of assessing the current workload and historical error rate curves of the manual picking team when a shortage of specific high-skilled human resources is identified, in order to determine the safe carrying capacity limit of the manual picking team, specifically includes:

[0132] In the manual sorting operation area, environmental sensing devices are deployed to obtain environmental information of the operation area in real time, including noise intensity, light level and temperature information.

[0133] By analyzing the collaborative behavior data of the manual picking team during the operation, the smoothness of internal team collaboration can be identified.

[0134] A situational awareness questionnaire was administered to the members of the manual sorting team to obtain their subjective perceptions of the current environment and teamwork.

[0135] By combining the environmental information, the smoothness of collaboration, and the subjective perception information, the potential impact of the members' work performance on the manual picking team can be inferred.

[0136] By combining the potential impact level with the current task load and the historical error rate curve, the safe carrying capacity limit of the manual picking team is dynamically adjusted.

[0137] Specifically, deploying environmental sensing devices in manual picking areas aims to monitor and quantify physical environmental factors affecting worker performance in real time. These devices may include, but are not limited to, noise sensors, light sensors, and temperature sensors. These sensors are strategically placed within the picking area to ensure comprehensive and real-time collection of information on noise intensity, light levels, and temperature. This environmental information is a crucial external factor influencing employee comfort, focus, and fatigue.

[0138] This study aims to quantify the efficiency and level of collaboration within a manual sorting team by analyzing their collaborative behavior data during operations. Collaborative behavior data can be derived from analyzing communication records between team members, task handover times, and the efficiency of collaborative task completion. For example, video analytics can be used to identify team members' movement trajectories and interaction patterns, or wearable devices can be used to monitor team members' physiological synchronicity, thereby identifying the smoothness of collaboration within the team. Collaboration smoothness is a key indicator reflecting the team's overall efficiency and resilience under pressure.

[0139] In practical applications, situational awareness questionnaires are administered to members of manual sorting teams to obtain their subjective feelings and perceptions about the current work environment and teamwork. These questionnaires can be designed periodically or event-triggered, gathering subjective information by asking employees about their feelings regarding noise, light, and temperature, as well as their evaluations of team communication, task allocation, and support levels. This subjective information supplements objective data, revealing employees' psychological state and satisfaction, thus providing a more comprehensive understanding of their work performance.

[0140] Furthermore, by combining the environmental information, the smoothness of collaboration, and the subjective perception information, the potential impact on the performance of members of the manual picking team can be inferred. For example, when noise levels are too high, lighting is insufficient, temperatures are uncomfortably low, team collaboration is slow, and employees subjectively experience high stress or low satisfaction, it can be inferred that their performance will be significantly negatively affected. This inference can be achieved through a pre-trained machine learning model that takes multi-dimensional data as input and outputs the probability or degree to which performance is affected.

[0141] Finally, the inferred potential impact level is combined with the current task load and the historical error rate curve to dynamically adjust the safety capacity limit of the manual picking team. This means that in harsh environments, poor collaboration, or when employees are subjectively dissatisfied, even if the current task load is not high, the team's safety capacity limit will be lowered to avoid errors and efficiency declines due to potential impacts. Conversely, in suitable environments, with efficient collaboration and good employee morale, the safety capacity limit may be appropriately increased.

[0142] The solution presented in this application effectively optimizes business processes because it overcomes the limitations of traditional methods that rely solely on task load and historical error rates when assessing the safe carrying capacity of manual picking teams. Through this technical solution, the application significantly enhances the refinement and intelligence of business process optimization. Compared to traditional methods that only consider task load and historical error rates, this application integrates multi-dimensional data, including environmental, collaborative, and subjective perception data, making the assessment of the safe carrying capacity of manual picking teams more accurate and dynamic. This effectively avoids decreased operational efficiency and increased error rates caused by hidden factors such as harsh environments, poor teamwork, or employee fatigue. The solution can adaptively adjust task allocation and business arrangements based on the team's real-time status and external context, thereby maximizing operational quality, reducing operational risks, and improving employee job satisfaction and health. This people-oriented dynamic optimization strategy not only improves the resilience and adaptability of the overall business process but also brings more significant economic and social benefits to the enterprise.

[0143] In some of the embodiments described above in this application, environmental sensing devices are proposed to be deployed in the manual picking area to obtain environmental information of the work area in real time. However, in practical applications, single or sparsely deployed environmental sensing devices may be difficult to comprehensively and accurately cover the entire work area, especially in complex picking environments, where there may be missing environmental information or local deviations, thereby affecting the accurate assessment of the performance of the manual picking team.

[0144] In this regard, this application further proposes the following steps for deploying environmental sensing devices in the manual picking operation area to obtain real-time environmental information of the operation area:

[0145] In the manual picking area, multiple environmental sensing devices are deployed, including noise sensors, light sensors, and temperature sensors. These environmental sensing devices form a distributed network to collect environmental information of the work area in real time.

[0146] The environmental information collected by the distributed network is fused with the micro-environmental information acquired by wearable environmental sensors configured on the manual picking team to construct an environmental information distribution map of the work area.

[0147] Based on the environmental information distribution map, identify areas where environmental information is missing or biased;

[0148] When an area with missing or deviated environmental information is identified, the missing or deviated environmental information is supplemented or corrected according to the environmental information distribution map.

[0149] Specifically, the deployment of multiple environmental sensing devices in the manual picking area, including noise sensors, light sensors, and temperature sensors, forms a distributed network to collect environmental information from the work area in real time. This means strategically arranging multiple sensor nodes within the manual picking area. These nodes work collaboratively to form a widely covered distributed network. Each environmental sensing device is configured to include at least a noise sensor, a light sensor, and a temperature sensor to comprehensively capture key environmental parameters within the work area. This distributed deployment enables real-time, multi-point collection of environmental information from the work area, resulting in more detailed and comprehensive environmental data.

[0150] The process of fusing environmental information collected by the distributed network with micro-environmental information acquired by wearable environmental sensors deployed on the manual picking team to construct an environmental information distribution map of the work area can be understood as follows: in addition to the fixedly deployed distributed network, wearable environmental sensors are also provided to members of the manual picking team. These wearable sensors can acquire micro-environmental information of individual team members, such as the noise intensity, light level, and temperature information they directly perceive. By spatiotemporally aligning and fusing the macro-environmental data collected by the distributed network with the micro-environmental data acquired by the wearable sensors, a more refined and individualized environmental information distribution map of the work area can be generated. This distribution map not only reflects the overall environmental condition but also reveals the specific environmental conditions faced by different locations and individuals.

[0151] In practical applications, identifying areas with missing or biased environmental information based on the environmental information distribution map refers to analyzing the constructed environmental information distribution map, for example, using interpolation algorithms, anomaly detection algorithms, or comparing it with historical data, to discover blank areas (missing data) or data points (biases) that do not match the actual situation. The purpose is to ensure the completeness and accuracy of environmental data.

[0152] Furthermore, the step of supplementing or correcting the missing or biased environmental information based on the environmental information distribution map when an area with missing or biased environmental information is identified means that once a missing or biased environmental information is identified, the system will use known data points in the environmental information distribution map to supplement the missing data through spatial interpolation, time series prediction, or other data filling techniques. For biased data, corrections can be made based on data from surrounding environmental sensing devices or historical trends to improve the reliability of the environmental information.

[0153] This application's solution, through the deployment of a distributed environmental sensing device network, achieves wide-area coverage and real-time collection of environmental information in the manual picking operation area, effectively avoiding local data blind spots that may be caused by the deployment of a single sensor. Through the above technical solution, this application can significantly improve the comprehensiveness, accuracy, and real-time performance of environmental information collection in the manual picking operation area. Compared to basic solutions that rely solely on fixed deployment equipment, this application, by introducing wearable environmental sensors and performing multi-source data fusion, can more accurately reflect the actual micro-environment in which team members are located, thereby more accurately assessing the potential impact of environmental factors on operational performance. Furthermore, by identifying, supplementing, or correcting missing or biased environmental information, it effectively avoids assessment errors caused by data quality issues, ensuring that subsequent dynamic adjustments to the safety carrying capacity limit of the manual picking team are more scientific and reasonable, thereby optimizing the decision-making quality and efficiency of the overall business process.

[0154] In some of the above embodiments, multiple environmental sensing devices, including noise sensors, light sensors, and temperature sensors, are deployed in the manual picking area to form a distributed network and collect environmental information of the work area in real time. However, in practical applications, these environmental sensing devices may experience performance drift, calibration deviations, or even abnormal fluctuations over time, leading to inaccurate or unreliable environmental information. If these problems are not addressed, the environmental information distribution map constructed based on inaccurate environmental information will fail to accurately reflect the actual situation of the work area, thus affecting the accurate inference of the potential impact on the performance of the manual picking team. Ultimately, this may lead to deviations in the correction of the safety carrying capacity limit of the manual picking team, hindering effective optimization of business processes. Therefore, this application further proposes a method for adaptive calibration and anomaly detection of environmental sensing devices to ensure the accuracy and reliability of the collected environmental information.

[0155] In the manual picking area, multiple environmental sensing devices are deployed, including noise sensors, light sensors, and temperature sensors. These devices form a distributed network to collect real-time environmental information about the work area, including:

[0156] In each environmental sensing device in the distributed network, a self-test program is periodically triggered, which generates sensor response data through an internal reference source or a preset standard signal.

[0157] The sensor response data is compared with the historical calibration data of the environmental sensing device to identify whether the environmental sensing device has performance drift or calibration deviation.

[0158] When the performance drift or calibration deviation is detected, the data output of the environmental sensing device is adjusted according to the degree of the performance drift or calibration deviation in order to perform adaptive calibration.

[0159] The system continuously monitors the environmental information collected by the environmental sensing device. When the environmental information shows abnormal fluctuations that are significantly inconsistent with the data of surrounding devices or historical trends, the abnormal detection mechanism is triggered and the abnormal fluctuations are marked.

[0160] Specifically, to ensure the accuracy and reliability of the data collected by each environmental sensing device in the distributed network, this application proposes a series of optimization measures. First, a self-test program is periodically triggered on each environmental sensing device. This self-test program aims to simulate specific environmental conditions by utilizing an internal reference source or a preset standard signal, and generate corresponding sensor response data. For example, a noise sensor can play a sound signal of known frequency and intensity, a light sensor can activate an internal standard light source, and a temperature sensor can use internal heating or cooling elements to reach a reference temperature. The sensors' responses to these standard inputs are then recorded.

[0161] Furthermore, the generated sensor response data is compared with the historical calibration data of the environmental sensing device. Historical calibration data is typically baseline data recorded after calibration with precision instruments at the time of manufacture or during periodic maintenance, reflecting the device's response characteristics under ideal conditions. This comparison identifies the differences between the current sensor response data and the historical baseline, thereby determining whether the environmental sensing device exhibits performance drift (i.e., slow changes in sensor output over time) or calibration deviation (i.e., systematic errors between the sensor output and the true value).

[0162] When performance drift or calibration deviation is detected, the system adjusts the data output of the environmental sensing device according to the degree of drift or deviation to achieve adaptive calibration. This means that if the sensor tends to overestimate or underestimate a certain environmental parameter, its output data will be corrected in real time to make it closer to the true value. This adaptive calibration mechanism can dynamically compensate for the effects of sensor aging or environmental changes, thereby ensuring the continuous accuracy of the data.

[0163] In addition, the system continuously monitors environmental information collected by environmental sensing devices. When environmental information collected by a device shows abnormal fluctuations that are significantly inconsistent with data from other surrounding devices or historical trends—for example, a sudden and abnormal rise or fall in temperature in a certain area, while the surrounding areas or historical data do not show such drastic changes—the system will trigger an anomaly detection mechanism. Once such abnormal fluctuations are detected, the abnormal data will be immediately marked so that subsequent data processing and analysis can identify and properly handle these data anomalies that may be caused by sensor malfunctions, external interference, or extreme events, thus preventing them from misleading the overall environmental assessment.

[0164] This application's solution effectively addresses the issue of inaccurate or unreliable sensor data that may arise during the long-term operation of distributed environmental sensing networks by introducing periodic self-checks, historical data comparison, adaptive calibration, and anomaly detection mechanisms. Through these technical solutions, this application significantly improves the accuracy and reliability of environmental information collection in manual picking areas. Compared to basic solutions that only deploy environmental sensing devices for data collection, this application, through periodic self-checks and comparisons with historical calibration data, can promptly detect and correct sensor performance drift and calibration deviations, ensuring the long-term accuracy of environmental data. Furthermore, the introduction of anomaly detection mechanisms effectively avoids misleading overall assessments due to data anomalies caused by sensor malfunctions or external interference, thereby improving the authenticity and effectiveness of the environmental information distribution map. This high-precision, high-reliability environmental information allows for more accurate inferences about the potential impact on the performance of manual picking team members, enabling more precise correction of the safety capacity limits of the manual picking team. This provides a more reliable basis for subsequent order prioritization, picking task allocation, and delivery task scheduling, ultimately achieving more efficient and safer business process optimization.

[0165] Specifically, fusing the environmental information collected by the distributed network with the micro-environmental information acquired by wearable environmental sensors configured on the manual picking team to construct an environmental information distribution map of the work area may include the following steps:

[0166] Extract the first timestamp and first spatial identifier of the environmental information collected by the distributed network, and extract the second timestamp and second spatial identifier of the microenvironment information acquired by the wearable environmental sensor;

[0167] The first timestamp and the second timestamp are compared. When the difference between the compared timestamps exceeds a preset first threshold, the first timestamp and the second timestamp are calibrated.

[0168] The first spatial identifier is compared with the second spatial identifier. When the difference between the compared spatial identifiers exceeds a preset second threshold, the first spatial identifier and the second spatial identifier are calibrated.

[0169] Based on the calibrated timestamp and spatial identifier, the environmental information collected by the distributed network and the micro-environment information obtained by the wearable environmental sensor are mapped to a unified grid cell, and the information within each grid cell is aggregated to construct an environmental information distribution map of the work area.

[0170] The first and second timestamps refer to the data generation or acquisition times corresponding to environmental information collected by the distributed network and micro-environment information acquired by wearable environmental sensors, respectively. The first and second spatial identifiers refer to the geographical locations or spatial coordinates corresponding to this information. These timestamps and spatial identifiers are fundamental to ensuring the accurate fusion of data from different sources.

[0171] Furthermore, the first timestamp and the second timestamp are compared. When the difference between the compared timestamps exceeds a preset first threshold, the first and second timestamps are calibrated. The purpose is to eliminate potential time synchronization errors between different sensors or systems. For example, calibration can be performed using time synchronization protocols or time series alignment algorithms based on data characteristics to ensure the consistency of all data in the time dimension.

[0172] Furthermore, the first spatial identifier is compared with the second spatial identifier. When the difference between the compared spatial identifiers exceeds a preset second threshold, the first and second spatial identifiers are calibrated. The purpose of this calibration is to correct spatial deviations caused by different sensor positioning systems or deployment errors. For example, calibration can be performed through GPS data correction, fusion positioning of indoor positioning systems, or spatial transformation algorithms based on known reference points to ensure accurate spatial correspondence of all data.

[0173] Therefore, based on the calibrated timestamps and spatial identifiers, environmental information collected by the distributed network and micro-environmental information acquired by wearable environmental sensors are mapped to unified grid cells. Information within each grid cell is then aggregated to construct an environmental information distribution map of the work area. Specifically, the work area can be divided into a series of discrete grid cells, each representing a specific spatial region. Environmental data from the distributed network and wearable sensors, after time and spatial calibration, are assigned to their corresponding grid cells. Within each grid cell, the collected multi-source environmental information can be aggregated, such as by calculating averages, weighted averages, or performing data fusion algorithms, to generate comprehensive environmental information for that grid cell, ultimately forming an environmental information distribution map of the entire work area.

[0174] The proposed solution employs rigorous timestamp and spatial identification calibration of environmental information collected by a distributed network and micro-environmental information acquired by wearable environmental sensors, ensuring high consistency across time and space for multi-source heterogeneous data. This precise calibration enables subsequent mapping of these data to a unified grid cell and aggregation processing, overcoming the limitations of a single data source and avoiding data inaccuracies or information conflicts caused by temporal or spatial misalignments. This lays the foundation for constructing a high-precision, high-reliability distribution map of the operational area's environmental information.

[0175] The above technical solutions effectively address the common temporal and spatial inconsistencies in multi-source environmental data fusion, significantly improving the accuracy and reliability of environmental information distribution maps. This precise data fusion and calibration mechanism enables the constructed environmental information distribution maps to more realistically and meticulously reflect the actual environmental conditions of the work area, providing more solid and accurate data support for subsequent assessments of the potential impact of manual picking team performance, thereby enhancing the effectiveness and relevance of business process optimization strategies.

[0176] refer to Figure 3 , Figure 3 This is a schematic diagram of the structure of a business process optimization system based on artificial intelligence big data algorithms provided in an embodiment of the present invention, including:

[0177] The input end is used to acquire business operation data and analyze the business operation data to identify combined behaviors that reflect specific operational anomalies.

[0178] The parsing end is used to respond to the combined performance, obtain external context information, and parse the external context information to identify external events related to the specific operational anomaly.

[0179] The generation end is used to establish a causal relationship between the external event and the specific operational anomaly; and to generate an optimization strategy for the external event based on the causal relationship.

[0180] This application's business process optimization system, through the collaborative work of its input, parsing, and generation ends, overcomes the limitations of traditional systems in deeply understanding the underlying causes and generating effective optimization strategies when faced with complex operational anomalies triggered by external unforeseen events. The system performs in-depth analysis of business operation data through the input end, identifying combinations of behaviors reflecting specific operational anomalies. Subsequently, the parsing end responds to these behaviors, proactively acquiring and parsing external contextual information to identify external events related to the operational anomalies. Finally, the generation end establishes causal relationships between external events and operational anomalies based on this information, and generates targeted and fundamental optimization strategies accordingly. Therefore, this system can fundamentally solve problems in business processes, improve operational efficiency and resilience, and avoid the shortcomings of traditional systems that only address the symptoms, not the root causes.

[0181] The core of the business process optimization system in this application lies in the collaboration and integration of various functional modules.

[0182] Specifically, the input terminal's function is to acquire and analyze business operation data to identify combinations of behaviors reflecting specific operational anomalies. The detailed process of data acquisition and analysis involved in this function has already been described in the above embodiments and will not be repeated here. It is important to emphasize that the input terminal can be configured in various forms. For example, it can be a data acquisition module that connects to various internal business systems (such as enterprise resource planning systems, customer relationship management systems, supply chain management systems, etc.) through a preset data interface to periodically or in real-time retrieve data. Furthermore, the input terminal can also include a data preprocessing unit for cleaning, formatting, and standardizing the raw business operation data to ensure the accuracy of subsequent analysis.

[0183] Furthermore, regarding the parsing end, its function is to respond to combined performance, acquire and parse external contextual information to identify external events related to specific operational anomalies. The detailed process of acquiring and parsing external contextual information involved in this function has already been described in the above embodiments and will not be repeated here. It is important to emphasize that the parsing end can be implemented as a context-aware module, which can integrate multiple external data source interfaces, such as connecting to third-party weather services, traffic information platforms, news aggregators, etc., via application programming interfaces. The parsing end can also include an information processing unit for structuring the acquired heterogeneous external contextual information, such as using keyword matching, rule engines, or simple text classifiers to initially identify potential external events.

[0184] Furthermore, regarding the generation end, its function is to establish causal relationships between external events and specific operational anomalies, and to generate optimization strategies for external events based on these causal relationships. The detailed process of establishing causal relationships and generating optimization strategies involved in this function has already been described in the above embodiments, and will not be repeated here. It is important to emphasize that the generation end can be implemented as a decision support module, which may contain a causal model library storing predefined causal rules or simple statistical correlation models. In addition, the generation end may also contain a strategy generation unit, which can generate preliminary optimization suggestions based on a preset strategy template or simple decision tree logic, according to the identified causal relationships. For example, when a certain external event is identified as causing a specific operational anomaly, the generation end can select one or more matching strategies from a preset strategy list for output.

[0185] The business process optimization system presented in this application represents a significant advancement compared to existing business process optimization systems. Traditional systems often focus on analyzing internal operational data to optimize established processes, but their perception and decision-making capabilities are insufficient when faced with operational anomalies caused by external unforeseen events and involving complex causal relationships. For example, existing systems may only be able to detect a decline in order processing efficiency through internal data, but struggle to proactively identify and link it to deeper causes such as extreme weather or human resource shortages. The system in this application, through the collaborative design of its input, parsing, and generation ends, achieves deep perception of external contextual information and intelligent construction of causal relationships. This capability enables the system to fundamentally understand the causes of operational anomalies and generate more targeted and forward-looking optimization strategies, thereby effectively improving the resilience, adaptability, and overall efficiency of business processes, and avoiding the decision-making blind spots and superficial solutions that may occur in traditional systems under complex and dynamic environments.

[0186] The above description is merely an embodiment of this application and is not intended to limit the scope of protection of this application. Various modifications and variations can be made to this application by those skilled in the art. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of this application should be included within the scope of protection of this application.

Claims

1. A business process optimization method based on artificial intelligence big data algorithms, characterized in that, include: Acquire business operation data and analyze the business operation data to identify combined behaviors that reflect specific operational anomalies; In response to the combined performance, external context information is acquired and parsed to identify external events related to the specific operational anomaly; Establish a causal relationship between the external event and the specific operational anomaly; Based on the causal relationship, an optimization strategy for the external event is generated.

2. The business process optimization method based on artificial intelligence big data algorithms according to claim 1, characterized in that, In response to the combined performance, external context information is acquired and parsed to identify external events related to the specific operational anomaly, including: When delivery vehicles travel on specific transport routes, the wheels are periodically induced to slip slightly, and wheel speed data and vehicle acceleration data are collected. Based on the collected wheel speed data, vehicle acceleration data, and braking pulse parameters, the current road surface micro-friction index is inferred. The inferred micro-friction index is compared with a preset safety threshold to identify whether there is a hidden low-friction risk in the specific transport trunk line; When the hidden low-friction risk is identified, the risk information, along with the vehicle location, time, and external weather report, is sent to the business process optimization system to identify external events related to the operational anomaly.

3. The business process optimization method based on artificial intelligence big data algorithms according to claim 1, characterized in that, The optimization strategy for generating the external event based on the causal relationship includes: When a shortage of specific high-skilled human resources is identified, the current workload and historical error rate curves of the manual picking team are assessed to determine the safe carrying capacity limit of the manual picking team; When a hidden low-friction risk is identified on a specific transport route, a safe recommended speed is calculated and pushed to drivers on the affected road section, and the recommended safe following distance of the fleet is adjusted. Based on the safe carrying capacity limit of the manual picking team, the recommended safe vehicle speed, and the recommended safe distance of the fleet, the orders to be processed are prioritized. Based on the priority ranking results, adjust the allocation of picking tasks and the arrangement of delivery tasks.

4. The business process optimization method based on artificial intelligence big data algorithms according to claim 3, characterized in that, When a shortage of specific high-skilled human resources is identified, the current workload and historical error rate curves of the manual picking team are assessed to determine the safe carrying capacity limit of the manual picking team, including: Feature analysis is performed on picking tasks that need to be diverted to identify the complexity level and the type of operational skills required for the picking tasks; Query the skill profile and historical task performance data of the manual picking team to assess the real-time effective carrying capacity of the manual picking team when handling complex tasks; The real-time effective carrying capacity and historical error rate curve are dynamically adjusted to determine the safe carrying capacity limit of the manual picking team when handling the complex task.

5. The business process optimization method based on artificial intelligence big data algorithms according to claim 3, characterized in that, When a shortage of specific high-skilled human resources is identified, the current workload and historical error rate curves of the manual picking team are assessed to determine the safe carrying capacity limit of the manual picking team, including: The physiological state of the members of the manual sorting team is monitored in real time to obtain physiological indicators; A subjective fatigue survey was conducted on the members of the manual sorting team to obtain fatigue information; The physiological indicators and fatigue information are fused together to infer the real-time fatigue level of the members of the manual sorting team; By combining the real-time fatigue level, the current task load, and the historical error rate curve, the safe carrying capacity limit of the members of the manual picking team is adjusted.

6. The business process optimization method based on artificial intelligence big data algorithms according to claim 3, characterized in that, When a shortage of specific high-skilled human resources is identified, the current workload and historical error rate curves of the manual picking team are assessed to determine the safe carrying capacity limit of the manual picking team, including: In the manual sorting operation area, environmental sensing devices are deployed to obtain environmental information of the operation area in real time, including noise intensity, light level and temperature information. By analyzing the collaborative behavior data of the manual picking team during the operation, the smoothness of internal team collaboration can be identified. A situational awareness questionnaire was administered to the members of the manual sorting team to obtain their subjective perceptions of the current environment and teamwork. By combining the environmental information, the smoothness of collaboration, and the subjective perception information, the potential impact of the members' work performance on the manual picking team can be inferred. By combining the potential impact level with the current task load and the historical error rate curve, the safe carrying capacity limit of the manual picking team is dynamically adjusted.

7. The business process optimization method based on artificial intelligence big data algorithms according to claim 6, characterized in that, The deployment of environmental sensing devices in the manual picking area to acquire real-time environmental information includes: In the manual picking area, multiple environmental sensing devices are deployed, including noise sensors, light sensors, and temperature sensors. These environmental sensing devices form a distributed network to collect environmental information of the work area in real time. The environmental information collected by the distributed network is fused with the micro-environmental information acquired by wearable environmental sensors configured on the manual picking team to construct an environmental information distribution map of the work area. Based on the environmental information distribution map, identify areas where environmental information is missing or biased; When an area with missing or deviated environmental information is identified, the missing or deviated environmental information is supplemented or corrected according to the environmental information distribution map.

8. The business process optimization method based on artificial intelligence big data algorithms according to claim 7, characterized in that, In the manual picking area, multiple environmental sensing devices are deployed, including noise sensors, light sensors, and temperature sensors. These devices form a distributed network to collect real-time environmental information about the work area, including: In each environmental sensing device in the distributed network, a self-test program is periodically triggered, which generates sensor response data through an internal reference source or a preset standard signal. The sensor response data is compared with the historical calibration data of the environmental sensing device to identify whether the environmental sensing device has performance drift or calibration deviation. When the performance drift or calibration deviation is detected, the data output of the environmental sensing device is adjusted according to the degree of the performance drift or calibration deviation in order to perform adaptive calibration. The system continuously monitors the environmental information collected by the environmental sensing device. When the environmental information shows abnormal fluctuations that are significantly inconsistent with the data of surrounding devices or historical trends, the abnormal detection mechanism is triggered and the abnormal fluctuations are marked.

9. A business process optimization method based on artificial intelligence big data algorithms according to claim 7, characterized in that, The step of fusing the environmental information collected by the distributed network with wearable environmental sensors configured on the manual picking team to construct an environmental information distribution map of the work area includes: Extract the first timestamp and first spatial identifier of the environmental information collected by the distributed network, and extract the second timestamp and second spatial identifier of the microenvironment information acquired by the wearable environmental sensor; The first timestamp and the second timestamp are compared. When the difference between the compared timestamps exceeds a preset first threshold, the first timestamp and the second timestamp are calibrated. The first spatial identifier is compared with the second spatial identifier. When the difference between the compared spatial identifiers exceeds a preset second threshold, the first spatial identifier and the second spatial identifier are calibrated. Based on the calibrated timestamp and spatial identifier, the environmental information collected by the distributed network and the micro-environment information obtained by the wearable environmental sensor are mapped to a unified grid cell, and the information within each grid cell is aggregated to construct an environmental information distribution map of the work area.

10. A business process optimization system based on artificial intelligence big data algorithms, characterized in that, include: The input end is used to acquire business operation data and analyze the business operation data to identify combined behaviors that reflect specific operational anomalies. The parsing end is used to respond to the combined performance, obtain external context information, and parse the external context information to identify external events related to the specific operational anomaly. The generation end is used to establish a causal relationship between the external event and the specific operational anomaly; Based on the causal relationship, an optimization strategy for the external event is generated.

Images