A collaborative robot system for active grid marketing and service process thereof

Abstract

The application discloses a kind of collaborative robot system and service process for active power grid marketing, belong to artificial intelligence and electric power marketing service technical field.System includes: robot perception interaction end, cloud intelligent analysis hub and dynamic customer memory bank, robot perception interaction end deploys lightweight vision model, real-time identification customer emotion, body state and explicit feature, and initiatively trigger service;Cloud intelligent analysis hub uses power grid knowledge enhanced large language model, carries out depth demand analysis and individualized strategy generation to multi-modal data;Dynamic customer memory bank stores customer historical interaction record and portrait information in the form of knowledge graph, realizes the continuity and memory of service.The application is through edge small model+cloud big model collaborative architecture, general AI capability and power grid business knowledge are deeply fused, so that robot can actively identify customer demand, provide marketing service, improve customer experience and marketing conversion rate of electric power business hall.

Description

Technical Field

[0001] This invention belongs to the field of artificial intelligence and power marketing service technology, and particularly relates to a collaborative robot system and its service process for proactive power grid marketing. Background Technology

[0002] As power grid companies transform into energy internet service providers, the traditional passive marketing model of "waiting for customers to come to you" can no longer meet customers' demands for personalized, convenient, and forward-looking services. Currently, power grid marketing services mainly face the following technical challenges:

[0003] First, the pressure and efficiency bottleneck of manual services.

[0004] During peak hours, the large number of customers in the sales hall puts immense pressure on human operators, leading to slow service responses. Marketing staff struggle to gain a timely and comprehensive understanding of each customer's potential needs, resulting in low marketing conversion rates.

[0005] Second, the existing service robots lack sufficient intelligence.

[0006] Currently, most of the tour guide robots deployed in business halls are based on fixed scripts and can only perform preset guidance or query functions. They lack deep semantic understanding capabilities, cannot personalize interactions based on customers' emotions, identity characteristics, and context, and cannot form a continuous understanding of customers, resulting in a rigid service experience.

[0007] Third, customer data is fragmented and lacks effective utilization.

[0008] Customer offline interaction records (such as consultation history and behavioral characteristics) are disconnected from online data, failing to form a unified customer profile. Each service is "starting from scratch," making it impossible to provide consistent service based on historical memories, thus hindering the improvement of customer loyalty.

[0009] Fourth, the application of advanced AI technologies is costly. Deploying large language models with powerful understanding and generation capabilities directly onto mobile robots places extremely high demands on computing power, power consumption, and network latency, resulting in enormous deployment costs and difficulties. How to introduce the intelligent capabilities of large models into robot systems while ensuring real-time response and controlling hardware costs is a pressing technical challenge that needs to be addressed.

[0010] Therefore, developing an intelligent robot system with proactive marketing capabilities and the ability to provide personalized services at low cost has significant application value. Summary of the Invention

[0011] This invention addresses the problems of low efficiency of manual labor, insufficient intelligence of service robots, and lack of continuity and initiative in customer service in existing power grid marketing services. It proposes an intelligent robot system and its service process that utilizes a large-scale, small-scale model collaboration. This system deploys a lightweight visual small model on the robot body to achieve real-time perception of customer emotions, posture, and behavioral characteristics, triggering proactive services. Simultaneously, it leverages a large-scale model in the cloud to perform in-depth demand analysis and memory-based decision-making on multimodal data, generating personalized proactive marketing service strategies. The small-scale model ensures real-time response and privacy security, while the large-scale model in the cloud provides powerful semantic understanding and strategy generation capabilities. Working together, the two drive the robot to accurately and proactively complete a closed-loop marketing service of "perception-understanding-decision-action," significantly improving service efficiency and customer experience.

[0012] The solution adopted by this invention to solve its technical problem is as follows:

[0013] A collaborative robot system for proactive power grid marketing, comprising:

[0014] The robot's perception and interaction terminal is deployed on the robot body and is used for environmental perception, basic interaction and command execution in physical space;

[0015] The cloud-based intelligent analysis hub, deployed on a cloud server, is used to deeply understand customer needs and generate service strategies.

[0016] A dynamic customer memory database, deployed in the cloud or on a standalone database server, is used to store customer information and historical interaction records in the form of a knowledge graph.

[0017] The robot's perception and interaction terminal, cloud-based intelligent analysis center, and dynamic customer memory database work collaboratively through high-speed network communication to achieve a closed-loop marketing service encompassing perception, understanding, decision-making, and action.

[0018] The hardware platform of the robot's perception and interaction terminal includes: an autonomous navigation mobile chassis, a depth camera, a microphone array, a speaker, a touch screen, and an edge computing unit;

[0019] The edge computing unit is equipped with a lightweight visual recognition model, a voice activity detection module, and a motion control module.

[0020] The lightweight visual recognition model employs a multi-model collaboration scheme. After pruning and INT8 quantization, it is deployed on an edge computing unit. Its output structured perception results include: customer emotional state, body language, explicit features, and group relationships.

[0021] The cloud-based intelligent analysis hub includes: a memory retrieval and fusion module, a large language model module for power grid knowledge enhancement, a strategy generation module, and a task orchestration module;

[0022] The aforementioned large language model for enhancing power grid knowledge is derived from a general large model by adapting it to the domain using internal policy documents, business manuals, and marketing cases of power grid companies.

[0023] As a preferred embodiment of the present invention

[0024] The robot's sensory interaction terminal is configured to perform the following steps:

[0025] Step S101: Environmental inspection and lightweight feature perception.

[0026] The robot moves within the power business hall or community service station, and collects real-time video streams through the depth camera, with the lightweight visual recognition model outputting structured perception results.

[0027] Step S102: Active service trigger judgment,

[0028] The decision logic module within the edge computing unit judges the perception results according to preset rules, and generates an active service intent when the triggering conditions are met.

[0029] Step S103: Multimodal data acquisition and uploading.

[0030] The robot body navigates to the vicinity of the customer, collects voice and video data and structured perception tags, and encrypts and uploads them to the cloud-based intelligent analysis center;

[0031] Step S104: Receive cloud instructions and perform interaction.

[0032] The robot body executes voice broadcasts, displays on the screen, and performs physical actions according to the instructions returned by the cloud-based intelligent analysis center, and continuously collects customer feedback and transmits it back to the cloud.

[0033] As a preferred embodiment of the present invention

[0034] The conditions under which the cloud-based intelligent analysis center triggers proactive services include any of the following:

[0035] The conditions under which the cloud-based intelligent analysis center triggers proactive services include any of the following:

[0036] The customer exhibits help-seeking behavior (waving, repeatedly checking their watch, pacing back and forth) with a confidence level higher than 0.75;

[0037] The customer's posture is such that they stand and look for more than 8 seconds, and they are not in the queuing area or the service area;

[0038] The client's overt characteristics are that they are elderly and unaccompanied, and they also exhibit abnormal behavior.

[0039] Customers marked as needing assistance in the dynamic customer memory.

[0040] As a preferred embodiment of the present invention

[0041] The cloud-based intelligent analysis hub is configured to perform the following steps:

[0042] Step S201: Memory retrieval and context fusion.

[0043] Extract customer features, retrieve matching data from the dynamic customer memory database, and perform multimodal fusion of current perceived data and historical memory;

[0044] Step S202: In-depth requirements analysis and classification based on a large model.

[0045] The fused contextual information is input into the large language model that enhances power grid knowledge, and the output includes demand type, urgency level, intention score, and deep demand inference.

[0046] Step S203: Personalized strategy generation,

[0047] Based on the analysis results and historical information in step S202, a personalized proactive service strategy is generated.

[0048] As a preferred embodiment of the present invention

[0049] The dynamic customer memory bank is stored in the form of a knowledge graph, with customers as entity nodes. Each entity node contains attributes such as name, social role tags, personality traits, interests and hobbies, and investment intentions.

[0050] The entity node is associated with the following sub-nodes: visit event node, demand tag node, service record node, feedback evaluation node, and to-do tracking node.

[0051] As a preferred embodiment of the present invention

[0052] The dynamic client memory is configured with a memory update mechanism:

[0053] After each service interaction, the cloud-based intelligent analysis center will automatically extract key information from the dialogue log and generate a structured summary.

[0054] The structured summary includes: newly discovered demand tags, service results, and updates to customer attributes.

[0055] As a preferred embodiment of the present invention

[0056] The dynamic client memory is configured with active service triggering rules:

[0057] When a customer is tagged with a specific label and remains unreached for a preset period, a task reminder is generated. The robot then proactively triggers the service upon the customer's next visit.

[0058] The active service triggering rules are configured through the management interface and stored in the rule engine.

[0059] As a preferred embodiment of the present invention

[0060] The robot's perception and interaction terminals are set up in multiple groups. When multiple robots are deployed in a large business hall, the cloud-based intelligent analysis center maintains the global map and customer location information, enabling obstacle avoidance coordination and seamless handover of customer service between the robot's perception and interaction terminals.

[0061] As a preferred embodiment of the present invention

[0062] The system is also configured with an anomaly handling process: when the cloud is unresponsive or the network is interrupted, the robot edge terminal activates a degradation mode and completes basic services based on the locally cached rule base.

[0063] A service process for a collaborative robot system for proactive power grid marketing includes the following steps:

[0064] S1. The robot body performs environmental inspection and perception triggering, and outputs structured perception results;

[0065] S2. Determine whether to trigger active service according to preset rules. If so, collect multimodal data and upload it to the cloud-based intelligent analysis center.

[0066] S3. The cloud-based intelligent analysis center performs memory retrieval and context fusion, conducts in-depth demand analysis and classification through a large language model enhanced with power grid knowledge, generates personalized strategies and issues instructions.

[0067] The personalized strategies include: proactive greetings, guiding users to designated areas, displaying electricity consumption analysis reports, recommending electricity service packages, pushing information to associated users, and processing business orders.

[0068] S4. The robot body receives instructions and performs interactions, while continuously collecting customer feedback and transmitting it back to the cloud-based intelligent analysis center.

[0069] S5. After the service is completed, the dynamic customer memory bank is updated based on the key information extracted from the cloud-based big model.

[0070] S6. Subsequent proactive service steps: Based on the triggering rules in the dynamic customer memory bank, the system will proactively initiate follow-up services when the customer visits again, or push service information through external communication channels.

[0071] Compared with the prior art, the present invention has the following beneficial effects:

[0072] This invention discloses a collaborative robot system and its service process for proactive power grid marketing. Primarily used in power grid marketing scenarios such as power business halls and community service stations, it aims to achieve proactive customer identification, demand analysis, memory-based service, and precise marketing. It has the following significant advantages:

[0073] 1. Shift marketing services from "passive response" to "proactive insight":

[0074] Unlike existing tour guide robots that can only passively respond to fixed instructions, the system of this invention uses a lightweight edge model to identify customers' emotional state (such as confusion or anxiety), physical behavior (such as standing still or lingering), and explicit characteristics (such as elderly people being alone) in real time. Based on preset rules, it actively triggers service processes, enabling the robot to autonomously discover potential service recipients and intervene in advance, significantly improving service coverage and customer experience.

[0075] 2. Adopting a collaborative architecture of "small edge models paired with large cloud models" to balance real-time performance and intelligence:

[0076] Existing technologies face bottlenecks in computing power, power consumption, and latency when deploying large models directly on the robot body; while using only local small models results in insufficient semantic understanding capabilities. This system deploys a lightweight visual model at the robot's edge to ensure low latency and privacy security for customer feature recognition; and deploys a large language model enhanced with power grid knowledge in the cloud to provide in-depth requirements analysis, multi-turn dialogue generation, and personalized strategy output.

[0077] 4. Construct a "dynamic customer memory map" to achieve service continuity and personalization:

[0078] Existing technologies disconnect customer offline interaction records from online data, requiring each service to start from scratch. This system uses a knowledge graph to create a dynamic digital memory profile for each customer, storing information such as historical visits, demand tags, service records, personality traits, and investment intentions. This profile is automatically updated after each service using a cloud-based big data model. Based on this memory graph, the chatbot can provide consistent "customer recognition" services (such as proactively addressing customers and recalling historical preferences), and supports task tracking and proactive follow-ups, greatly enhancing customer loyalty and marketing precision.

[0079] 5. Deeply empower vertical scenarios of power grid marketing to improve business conversion capabilities:

[0080] Unlike general-purpose service robots, this system deeply integrates general AI capabilities with power grid business knowledge. By fine-tuning the large model using power grid policy documents, business manuals, and marketing cases, the robot can understand professional terms such as electricity pricing policies, photovoltaic promotion, energy efficiency services, and the electric vehicle ecosystem, and generate personalized marketing strategies that conform to business specifications (such as recommending peak-valley electricity pricing and assisting with children's automatic deductions). Attached Figure Description

[0081] Figure 1 This is an overall architecture diagram of a collaborative robot system for proactive power grid marketing proposed in this invention;

[0082] Figure 2 This is a flowchart illustrating the computation of the robot's perception and interaction terminal in a collaborative robot system for proactive power grid marketing, as proposed in this invention.

[0083] Figure 3 This is a flowchart of the cloud-based intelligent analysis center in a collaborative robot system for proactive power grid marketing proposed in this invention.

[0084] Figure 4 This is a schematic diagram of a dynamic customer memory map update mechanism in a collaborative robot system for proactive power grid marketing proposed in this invention. Detailed Implementation

[0085] The specific embodiments of the present invention are described below with reference to the accompanying drawings and examples:

[0086] It should be noted that the structures, colors, proportions, sizes, etc. shown in the accompanying drawings are only used to complement the content disclosed in the specification, so that those skilled in the art can understand and read them, and are not intended to limit the conditions under which the present invention can be implemented. Any modifications to the structure, changes in the proportions, or adjustments to the size, without affecting the effects and objectives that the present invention can produce, should still fall within the scope of the technical content disclosed in the present invention.

[0087] At the same time, in the description of the present invention, it should be understood that the terms "one end", "the other end", "middle", "upper", "side", "top", "inner", "front", "center", "both ends", etc., indicate the orientation or positional relationship based on the orientation or positional relationship shown in the drawings, and are only for the convenience of describing the present invention and simplifying the description, and do not indicate or imply that the device or element referred to must have a specific orientation, or be constructed and operated in a specific orientation, and therefore should not be construed as a limitation of the present invention.

[0088] Furthermore, the terms "first," "second," "third," and "fourth" are used for descriptive purposes only and should not be construed as indicating or implying relative importance or implicitly specifying the number of technical features indicated. Thus, a feature defined as "first," "second," "third," or "fourth" may explicitly or implicitly include at least one of those features.

[0089] In this invention, unless otherwise explicitly specified and limited, the terms "installation," "setting," "connection," "fixing," "screw connection," etc., should be interpreted broadly. For example, they can refer to a fixed connection, a detachable connection, or an integral part; they can refer to a mechanical connection or an electrical connection; they can refer to a direct connection or an indirect connection through an intermediate medium; they can refer to the internal connection of two components or the interaction between two components. Unless otherwise explicitly limited, those skilled in the art can understand the specific meaning of the above terms in this invention according to the specific circumstances.

[0090] like Figure 1 As shown, this invention provides a collaborative robot system for proactive power grid marketing. The system mainly includes: a robot perception and interaction terminal deployed on the robot body; a cloud-based intelligent analysis hub (i.e., a large model) deployed on a cloud server; and a dynamic customer memory database deployed in the cloud or on an independent database server. These three components work collaboratively through high-speed network communication to achieve a closed-loop marketing service encompassing "perception-understanding-decision-action".

[0091] The robot's perception and interaction terminal is responsible for environmental perception, basic interaction, and command execution in the physical space. Its hardware platform includes: an autonomous navigation mobile chassis, a depth camera, a microphone array, a speaker, a touchscreen display, and an edge computing unit. Where necessary, the edge computing unit deploys a lightweight visual recognition model, a voice activity detection module, and a motion control module.

[0092] like Figure 2 As shown, the service process of the robot's perception and interaction terminal specifically includes the following steps:

[0093] Step S101: Environmental inspection and lightweight feature perception.

[0094] The robot autonomously moves along a pre-set inspection path within power service halls or community service stations, or roams randomly. Its onboard depth camera captures real-time video streams at 15-30 frames per second, which are then input into a lightweight visual recognition model (i.e., a small model) deployed on an edge computing unit. This model employs a multi-model collaborative scheme (PicoDet-Lite detection backbone + EfficientFace-Lite sentiment analysis + LitePose pose estimation + ByteTrack multi-object tracking), and after pruning and INT8 quantization, is deployed on the edge computing unit. It can process one frame of image within 50ms and output structured perception results in real time.

[0095] The structured perception results output by the lightweight visual recognition model include:

[0096] Customer emotional state: The system comprehensively judges the emotion category and confidence level by recognizing facial expressions (such as happiness, confusion, and anxiety) and analyzing voice tone.

[0097] Body posture and behavior: Identify customer postures through key human body points, including pausing and looking around (staying for more than 5 seconds with eyes directed towards the business display area), loitering (walking back and forth in non-passage areas), and leaving quickly;

[0098] Explicit characteristics: age group (children, youth, middle-aged, elderly), whether or not a safety helmet is worn (for power construction site services), whether or not paper bills are carried, etc.

[0099] Group relationships: By using pedestrian re-identification and distance clustering, we can determine whether there are accompanying persons and the intensity of interaction between people.

[0100] Step S102: Active service trigger judgment.

[0101] The decision logic module within the edge computing unit judges the perception results according to preset rules.

[0102] The conditions under which a robot system triggers proactive services include, but are not limited to:

[0103] 1. The customer exhibits help-seeking behavior (waving, repeatedly checking their watch, pacing back and forth) with a confidence level higher than 0.75;

[0104] 2. The customer's posture is "standing and observing" for more than 8 seconds, and they are not in the queuing area or the service area;

[0105] 3. The client's obvious characteristics are "elderly" and they are unaccompanied, while also exhibiting abnormal behavior;

[0106] The customer is marked as a "user in need of help" in the dynamic customer memory (the memory needs to be queried in advance through facial feature vector matching, and this query can be performed asynchronously).

[0107] When any triggering condition is met, the system generates an active service intent and proceeds to the next step.

[0108] Step S103: Multimodal data acquisition and uploading.

[0109] The robot autonomously navigates to the vicinity of the target customer (approximately 1-1.5 meters away, maintaining a non-invasive social distance) and initiates multimodal data collection: it collects the customer's voice through a microphone array (if the customer does not speak, it waits for 3 seconds or proactively greets them to encourage them to speak); it continuously captures 10-30 frames of images and corresponding depth maps using a depth camera. After collection, the robot packages the aforementioned voice and video data, along with the structured perception tags extracted in step S101, and encrypts and uploads them to the cloud-based intelligent analysis center. Simultaneously, the robot locally caches a copy of the data for rapid response.

[0110] Step S104: Receive instructions from the cloud and perform the interaction.

[0111] The robot awaits interactive instructions from the cloud, synthesizes speech based on these instructions, and plays it through a speaker; simultaneously, it controls the movement of its chassis or the execution of physical actions by its robotic arm. During execution, the robot continuously collects customer feedback (voice and facial expressions) and transmits it back to the cloud in real time to support dynamic adjustments for multi-turn dialogues. The instructions sent from the cloud include: the text to be read, the content displayed on the screen (such as an electricity price comparison chart), and robot actions (such as guiding the robot to a rest area or delivering promotional leaflets).

[0112] Preferably, the lightweight visual recognition model can adopt a lightweight multi-task visual model (e.g., using PicoDet-Lite as a unified backbone, integrating emotion recognition, pose estimation and attribute analysis branches), or it can adopt the Efficient-Lite series network (using a unified backbone to achieve multi-task output).

[0113] As another preferred embodiment of the present invention, the robot perception and interaction terminal is equipped with multiple sets. When multiple robots are deployed in a large business hall, the cloud-based intelligent analysis center maintains the global map and customer location information to avoid mutual interference between robots and achieve seamless handover of customer service (for example, when a customer moves from robot A area to robot B area, the memory is automatically transferred).

[0114] The cloud-based intelligent analysis hub, deployed on the power company's private or dedicated cloud servers, serves as the system's "brain," responsible for deeply understanding customer needs and generating service strategies. It primarily comprises three core sub-modules: a memory retrieval and fusion module, a large language model module for enhancing power grid knowledge, and a strategy generation and task orchestration module.

[0115] like Figure 3 As shown, the service process of the cloud-based intelligent analysis center is as follows:

[0116] Step S201: Memory retrieval and context fusion.

[0117] After receiving the data packet uploaded by the robot, the cloud-based intelligent analysis center first extracts the customer's facial feature vector or voiceprint features as query keys, and then performs a rapid retrieval and matching in the dynamic customer memory database. If an existing customer profile is matched, its historical interaction records, needs preferences, service results, and to-do items are retrieved; if no match is found, a new initial profile for a new customer is created. Subsequently, the current perceived data (emotional tags, body language, and environmental location) is fused with historical memory in a multimodal manner to form contextual information containing the following fields:

[0118] Static characteristics: age, social role (inferred from historical service records, such as "company electrician" or "elderly resident"), and historical needs labels;

[0119] Dynamic characteristics: current mood, current behavior, whether or not specific items are being carried (such as bills, helmets);

[0120] Environmental characteristics: the area of ​​the business hall (payment area, information area, photovoltaic display area), the number of people currently queuing, etc.

[0121] Step S202: In-depth requirements analysis and classification based on a large model.

[0122] The fused contextual information from step S201 is input into a large language model enhanced with power grid knowledge. Based on a general large model, this model uses internal power grid enterprise policy documents (such as detailed rules for tiered electricity pricing for residential use and distributed photovoltaic grid connection procedures), business manuals (95598 FAQ database), and marketing cases (dialogue records of successful promotion of electric vehicle charging stations) for domain adaptation, outputting structured analysis results, including:

[0123] Types of needs: Business processing (such as "name change / ownership transfer", "capacity increase"), information consultation (such as "electricity bill inquiry", "photovoltaic subsidy policy"), complaints and suggestions, business opportunity discovery (such as "intending to install charging piles", "consulting on energy storage solutions");

[0124] Urgency level: High (e.g., customer anger or power outage issues), Medium, Low;

[0125] Intention rating: A floating-point number between 0 and 1, representing the probability that a customer will accept the recommended service;

[0126] Inferring underlying needs: For example, when a customer asks "Why is my electricity bill so high this month?", the underlying need might be "I want to know how to save electricity" or "I suspect my electricity meter is malfunctioning."

[0127] Step S203: Personalized strategy generation.

[0128] Based on the above analysis results and combined with historical information in the dynamic customer memory bank, the strategy generation module generates personalized proactive service strategies. For example: "Ms. Li (historical records show she is an elderly person living alone, has repeatedly inquired about offline payment methods, and her children's information has been authorized for association), is currently holding a paper bill and is emotionally confused. The corresponding proactive service strategy generated would be:"

[0129] 1. Proactively greet her and recognize her habits (Aunt Li, are you here to pay the electricity bill again? Let me take a look at your bill and check your electricity usage this month.);

[0130] 2. Guide them to the rest area and simultaneously retrieve their household electricity consumption summary report;

[0131] 3. Use a large screen to clearly display the electricity consumption trends over the past three months, and explain peak hours in plain language;

[0132] 4. Proactively suggest and assist with the 'children's direct debit' or 'voice broadcast bill' service. With their consent, the monthly electricity bill details and electricity usage suggestions can be pushed to their children's mobile phones with one click.

[0133] 5. Based on their focus on 'energy saving', we recommend 'peak-valley electricity pricing' packages and provide a complimentary energy-saving tips card.

[0134] The dynamic customer memory bank is the "memory center" of the robot system. It uses customers as nodes and edges to represent relationships or events, and is updated according to the service situation.

[0135] like Figure 4 As shown, the data structure of the dynamic customer memory bank is stored in the form of a knowledge graph, with customers as entity nodes. Each entity node contains attributes: name, social role tags (such as "resident", "corporate representative", "elderly"), personality traits (derived through historical dialogue analysis, such as "patient" and "irritable"), hobbies (such as "interested in new energy vehicles"), and investment intentions (such as "photovoltaic intention degree 0.8").

[0136] Additionally, the client node is associated with the following edges and child nodes:

[0137] Visit event nodes: Record the time, duration, and reception robot ID for each visit;

[0138] Demand tag node: Records the demands identified in each interaction (such as "check electricity bill" or "inquire about photovoltaics"), with timestamps and weights;

[0139] Service record node: Records the service content and service results (success / failure / pending follow-up) provided by the robot or human agent;

[0140] Feedback and evaluation points: customer satisfaction ratings and written feedback;

[0141] To-do tracking nodes: Record items that require follow-up (such as "follow up on photovoltaic intentions in two weeks").

[0142] like Figure 4 As shown, the dynamic customer memory bank's memory update mechanism is as follows: After each service interaction, the cloud-based big data model automatically extracts key information from the dialogue log, generates a structured summary, and extracts the key information. The specific process is as follows:

[0143] The model extracts newly discovered demand tags (for example, if a customer inquires about the "electric vehicle charging pile installation process", then the tag "interested in charging piles" is added, with a weight of +1).

[0144] If the model extracts service results (e.g., "PV data has been pushed to customer's WeChat", then update the service record node);

[0145] Update customer attributes in the model (e.g., if the customer is determined to be "price-sensitive" based on this conversation, add the attribute).

[0146] The proactive service triggering rules for the dynamic customer memory database are as follows: the system allows marketers to configure triggering rules through the management interface, and the rules are stored in the rule engine. For example, when a customer is marked as a "high-intent photovoltaic customer" and has not been contacted again for more than two weeks, a task reminder can be generated. On the customer's next visit, the robot will recognize the customer and proactively approach them to inquire and follow up with services. These tasks will be inserted into the to-do queue, and when the robot detects that the customer has re-entered the business hall through facial recognition, proactive service will be triggered immediately.

[0147] The management interface refers to a graphical configuration platform for system maintenance personnel or power grid marketing managers. This interface is typically presented as a web application, accessed through a browser, and connected to the backend rule engine. Managers can create, modify, and delete trigger rules on this interface using methods such as clicking, entering thresholds, and dragging and dropping condition components, without modifying the underlying program code. For example, managers can set the rule condition: "Investment Intention.PV > 0.7 AND Last Contact Time > 14 days," and configure the corresponding action: "Generate Task Reminder, Mark Priority as High." After configuration, the system automatically compiles and loads the rule into the rule engine, making it effective in real time. The management interface also provides functions such as rule simulation testing, trigger log viewing, and rule conflict detection, ensuring flexible configuration and dynamic adjustment of marketing strategies. Through this management interface, power grid companies can update proactive service strategies at any time according to business needs without downtime or system redeployment, significantly improving the responsiveness and precision of marketing activities.

[0148] Preferably, the service process of a collaborative robot system for proactive power grid marketing can also be configured with an exception handling procedure: if there is no response from the cloud or the network is interrupted, the robot edge can enable a degradation mode, complete the basic service based on the locally cached rule base (such as simple question-and-answer pairs), and prompt the customer "the network is busy, we will follow up for you later".

[0149] A service process for a collaborative robot system for proactive power grid marketing includes the following steps:

[0150] S1. The robot body performs environmental surveys and lightweight feature perception, and outputs structured perception results;

[0151] S2. Determine whether to trigger active service based on preset rules. If so, collect multimodal data and upload it to the cloud-based intelligent analysis center.

[0152] S3, the cloud-based intelligent analysis center performs memory retrieval and context fusion, and uses a large language model enhanced with power grid knowledge to perform in-depth demand analysis and classification, generate personalized strategies and issue instructions;

[0153] Personalized strategies include: proactively greeting customers, guiding them to designated areas, displaying electricity usage analysis reports, recommending electricity service packages, conducting electricity safety mini-lectures (covering peak-valley electricity reminders and tiered pricing conversions), pushing information to associated users, and processing business work orders.

[0154] S4. The robot receives instructions and executes the interaction, while continuously collecting customer feedback and sending it back to the cloud-based intelligent analysis center.

[0155] S5. After the service is completed, the dynamic customer memory bank is updated based on the key information extracted from the cloud-based big model.

[0156] S6. Follow-up proactive service steps: Based on the triggering rules in the dynamic customer memory bank, the system will proactively initiate follow-up services when the customer visits again, or push service information through external communication channels.

[0157] A complete service process example of a collaborative robot system for proactive grid marketing (taking elderly customers living alone as an example).

[0158] Scene: At a power company service center, Ms. Li (68 years old, living alone, history shows she comes to the counter to pay her bill monthly and has complained about the complexity of online procedures) enters the hall holding a paper bill, looking confused. The specific service process includes:

[0159] 1. Robot Active Recognition: The robot recognizes Ms. Li's face through a camera, extracts features using an edge model, and asynchronously queries its memory, returning "long-time customer, elderly living alone, historical preference for offline payments." Simultaneously, the model identifies her emotion as "confused," and she is holding a paper bill. This triggers proactive service.

[0160] 2. Robot Greeter: The robot navigates to a point 1 meter in front of Ms. Li and announces based on a pre-generated greeting template (combined with historical addressing habits): "Aunt Li, you've arrived! Let me help you check the bill."

[0161] 3. Cloud-based in-depth analysis: The robot uploads multimodal data from the scene. The cloud search revealed that Ms. Li's children had authorized the association of her mobile phone number, and that she had complained in past conversations about "running the air conditioner a lot at night, resulting in high electricity bills." The large-scale model analysis identified the demand type as "information consultation + business opportunity discovery," with the deeper demand being "understanding energy-saving methods and reducing errands."

[0162] 4. Strategy Generation and Execution: Personalized strategies, as described above, are generated in the cloud. The robot guides Ms. Li to the large screen in the rest area, displays her electricity usage curve for the past three months, and explains in voice: "Auntie, look at this peak time at night. If we switch to off-peak pricing, we can save more than 30 yuan a month." Then it asks: "How about I help you set up off-peak pricing and link it to your son's mobile phone so he can check your electricity bill for you?" After Ms. Li agrees, the robot guides her to complete facial authorization through the large screen and automatically generates a work order, which is then pushed to the green channel at the counter.

[0163] 5. Memory Update: After the service is completed, the cloud will store the interaction record in the database: add the request tag "Apply for Peak-Valley Electricity Pricing", update the service record to "Applied", and add a follow-up tracking message "Follow-up visit three months later to check energy-saving effect". At the same time, add a "Trusted Robot Interaction" tag to the customer attributes.

[0164] 6. Proactive follow-up service: Two weeks later, the system detected the release of a new tiered electricity pricing policy in the region and automatically generated a push notification task to inform Ms. Li via chatbot or SMS, further enhancing customer loyalty.

[0165] The preferred embodiments of the present invention have been described in detail above with reference to the accompanying drawings. However, the present invention is not limited to the above embodiments. Within the scope of knowledge possessed by those skilled in the art, various changes can be made without departing from the spirit of the present invention.

[0166] Many other changes and modifications can be made without departing from the concept and scope of this invention. It should be understood that this invention is not limited to the specific embodiments, and the scope of this invention is defined by the appended claims.

Claims

1. A collaborative robot system for proactive power grid marketing, Its features are, Include: The robot's perception and interaction terminal is deployed on the robot body and is used for environmental perception, basic interaction and command execution in physical space; The cloud-based intelligent analysis hub, deployed on a cloud server, is used to deeply understand customer needs and generate service strategies. A dynamic customer memory database, deployed in the cloud or on a standalone database server, is used to store customer information and historical interaction records in the form of a knowledge graph. The robot's perception and interaction terminal, cloud-based intelligent analysis center, and dynamic customer memory bank work together through high-speed network communication to achieve a closed-loop marketing service of perception-understanding-decision-action. The hardware platform of the robot's perception and interaction terminal includes: an autonomous navigation mobile chassis, a depth camera, a microphone array, a speaker, a touch screen, and an edge computing unit; The edge computing unit is equipped with a lightweight visual recognition model, a voice activity detection module, and a motion control module. The lightweight visual recognition model adopts a multi-model collaboration scheme. After pruning and INT8 quantization, it is deployed on the edge computing unit. Its output structured perception results include: customer emotional state, body behavior, explicit features, and group relationships. The cloud-based intelligent analysis hub includes: a memory retrieval and fusion module, a large language model module for power grid knowledge enhancement, a strategy generation module, and a task orchestration module; The large language model for enhancing power grid knowledge is derived from a general large model by adapting it to the domain using internal policy documents, business manuals, and marketing cases of power grid companies.

2. A collaborative robot system for proactive power grid marketing as described in claim 1, Its features are, The robot's sensory interaction terminal is configured to perform the following steps: Step S101: Environmental inspection and lightweight feature perception. The robot moves within the power business hall or community service station, and collects real-time video streams through the depth camera, with the lightweight visual recognition model outputting structured perception results. Step S102: Active service trigger judgment, The decision logic module within the edge computing unit judges the perception results according to preset rules, and generates an active service intent when the triggering conditions are met. Step S103: Multimodal data acquisition and uploading. The robot body navigates to the vicinity of the customer, collects voice and video data and structured perception tags, and encrypts and uploads them to the cloud-based intelligent analysis center; Step S104: Receive cloud instructions and perform interaction. The robot body executes voice broadcasts, displays on the screen, and performs physical actions according to the instructions returned by the cloud-based intelligent analysis center, and continuously collects customer feedback and transmits it back to the cloud.

3. A collaborative robot system for proactive power grid marketing as described in claim 2, Its features are, The conditions under which the cloud-based intelligent analysis center triggers proactive services include any of the following: The customer seeks help and the confidence level is higher than 0.75; The customer's posture is such that they stand and look for more than 8 seconds, and they are not in the queuing area or the service area; The client's overt characteristics are that they are elderly and unaccompanied, and they also exhibit abnormal behavior. Customers marked as needing assistance in the dynamic customer memory.

4. A collaborative robot system for proactive power grid marketing as described in claim 1. Its features are, The cloud-based intelligent analysis hub is configured to perform the following steps: Step S201: Memory retrieval and context fusion. Extract customer features, retrieve matching data from the dynamic customer memory bank, and perform multimodal fusion of current perceived data and historical memory; Step S202: In-depth requirements analysis and classification based on a large model. The fused contextual information is input into the large language model that enhances power grid knowledge, and the output includes demand type, urgency level, intention score, and deep demand inference. Step S203: Personalized strategy generation, Based on the analysis results and historical information in step S202, a personalized proactive service strategy is generated.

5. A collaborative robot system for proactive power grid marketing as described in claim 4. Its features are, The dynamic customer memory bank is stored in the form of a knowledge graph, with customers as entity nodes. Each entity node contains attributes such as name, social role tags, personality traits, interests and hobbies, and investment intentions. The entity node is associated with the following sub-nodes: visit event node, demand tag node, service record node, feedback evaluation node, and to-do tracking node.

6. A collaborative robot system for proactive power grid marketing as described in claim 1, Its features are, The dynamic client memory is configured with a memory update mechanism: After each service interaction, the cloud-based intelligent analysis center will automatically extract key information from the dialogue log and generate a structured summary. The structured summary includes: newly discovered demand tags, service results, and updates to customer attributes.

7. A collaborative robot system for proactive power grid marketing as described in claim 1, Its features are, The dynamic client memory is configured with active service triggering rules: When a customer is tagged with a specific label and remains unreached for a preset period, a task reminder is generated. The robot then proactively triggers the service upon the customer's next visit. The active service triggering rules are configured through the management interface and stored in the rule engine.

8. A collaborative robot system for proactive power grid marketing as described in claim 1, Its features are, The robot's perception and interaction terminals are set up in multiple groups. When multiple robots are deployed in a large business hall, the cloud-based intelligent analysis center maintains the global map and customer location information, enabling obstacle avoidance coordination and seamless handover of customer service between the robot's perception and interaction terminals.

9. A collaborative robot system for proactive power grid marketing as described in claim 1. Its features are, The system is also configured with an anomaly handling process: when the cloud is unresponsive or the network is interrupted, the robot edge terminal activates a degradation mode and completes basic services based on the locally cached rule base.

10. A service process for a collaborative robot system for proactive power grid marketing, employing the collaborative robot system for proactive power grid marketing as described in any one of claims 1-9. Its features are, Includes the following steps: S1. The robot body performs environmental surveying and lightweight feature perception, and outputs structured perception results; S2. Determine whether to trigger active service according to preset rules. If so, collect multimodal data and upload it to the cloud-based intelligent analysis center. S3. The cloud-based intelligent analysis center performs memory retrieval and context fusion, conducts in-depth demand analysis and classification through a large language model enhanced with power grid knowledge, generates personalized strategies and issues instructions. The personalized strategies include: proactive greetings, guiding users to designated areas, displaying electricity consumption analysis reports, recommending electricity service packages and electricity safety mini-lessons, pushing information to associated users, and processing business work orders. S4. The robot body receives instructions and performs interactions, while continuously collecting customer feedback and transmitting it back to the cloud-based intelligent analysis center. S5. After the service is completed, the dynamic customer memory bank is updated based on the key information extracted from the cloud-based big model. S6. Subsequent proactive service steps: Based on the triggering rules in the dynamic customer memory bank, the system will proactively initiate follow-up services when the customer visits again, or push service information through external communication channels.

Images