CN-121997261-A - Intelligent robot knowledge base management method and system

CN121997261ACN 121997261 ACN121997261 ACN 121997261ACN-121997261-A

Abstract

The invention relates to the technical field of intelligent robot knowledge management, in particular to an intelligent robot knowledge base management method and system, comprising the following steps: and inputting the original data stream generated by the operation of the robot into a knowledge deposition module to deposit and layer, so as to form an initial knowledge deposition layer with a hierarchical structure. The initial knowledge deposit layer is imported into a knowledge smelting engine, which performs cross-layer fusion and recrystallization processing on the knowledge deposit layer to generate a standardized knowledge crystal set. The knowledge crystal set is conveyed to a topology braiding machine, and the knowledge crystal is subjected to space relation mapping and link braiding through the braiding machine to construct a dynamic knowledge topology network. And performing knowledge state evaluation and evolution tracking based on the dynamic knowledge topology network. According to the invention, the cognitive depth and the dynamic evolution capability of the knowledge base are improved by forming the structured knowledge unit and constructing the self-adaptive relation network.

Inventors

LI XIANGYANG
LI AILIN
Yao Qiuhui

Assignees

中科研(北京)科技发展中心

Dates

Publication Date: 20260508
Application Date: 20260126

Claims (10)

1. An intelligent robot knowledge base management method, which is characterized by comprising the following steps: inputting an original operation data stream generated in the operation process of the intelligent robot to a knowledge deposition module; depositing and layering the original operation data stream through the knowledge deposition module to form an initial knowledge deposition layer with a hierarchical structure; introducing the initial knowledge deposit layer into a knowledge smelting engine, and performing cross-layer fusion and recrystallization on the initial knowledge deposit layer by the knowledge smelting engine to generate a standardized knowledge crystal set; Delivering the knowledge crystal set to a topology braiding machine; carrying out space relation mapping and link braiding on knowledge crystals in the knowledge crystal set through a topology braiding machine to construct a dynamic knowledge topology network; and performing knowledge state evaluation and evolution tracking based on the dynamic knowledge topology network.
2. The method for intelligent robotic knowledge base management as set forth in claim 1, wherein depositing and layering the original operation data stream by the knowledge deposition module to form an initial knowledge deposition layer having a hierarchical structure comprises: The knowledge deposition module receives the original operation data stream and performs timeliness marking and source classification on the original operation data stream; Performing density analysis and event slicing on the classified original operation data stream, and extracting key event fragments from the data stream; depositing different key event fragments into different data levels according to semantic density and time continuity of the key event fragments; performing in-layer calibration and time stamp alignment on key event fragments deposited into each data level; Integrating all the calibrated and aligned data levels, and outputting the initial knowledge deposit layer with the hierarchical structure.
3. The intelligent robotic knowledge base management method of claim 1, wherein the step of cross-layer fusing and recrystallizing the initial knowledge deposit layer by the knowledge smelting engine to generate a standardized knowledge crystal set comprises: The knowledge smelting engine analyzes the hierarchical structure of the initial knowledge deposit layer and reads key event fragments in each hierarchy; Searching key event fragments with semantic association or causal association between different levels of the initial knowledge deposition layer; performing content fusion and context splicing on the found key event fragments with the association to form a preliminary knowledge block; Carrying out structural regularity and boundary definition on the preliminary knowledge block, and eliminating contradiction and redundant information inside the knowledge block; packaging the knowledge blocks which are completed to be regular and defined into knowledge crystals with uniform interface formats; and collecting all the packaged knowledge crystals to form the standardized knowledge crystal set.
4. The intelligent robot knowledge base management method of claim 1, wherein constructing a dynamic knowledge topology network by mapping and linking knowledge crystals in a knowledge crystal set by a topology braiding machine comprises: the topology braiding machine acquires the knowledge crystal set and analyzes the core semantic label and the attribute field of each knowledge crystal; calculating the association weight between the knowledge crystals based on the similarity of the core semantic tags and the complementarity of the attribute fields between any two knowledge crystals; according to the association weights between every two knowledge crystals, an initial relation diagram taking the knowledge crystals as nodes and the association weights as edge weights is constructed; introducing time dimension constraint in the initial relation diagram, and carrying out dynamic attenuation and enhancement operation on the edge weight to enable the relation diagram to have time sequence dynamics; and carrying out path optimization and redundant edge trimming on the relation graph with time sequence dynamics, and finally outputting the dynamic knowledge topology network.
5. The intelligent robot knowledge base management method as set forth in claim 4, wherein calculating the association weight between knowledge crystals based on the similarity of the core semantic tags and the complementarity of the attribute fields comprises: Extracting core semantic tag sets of the first knowledge crystal and the second knowledge crystal; calculating Jacquard similarity between the core semantic tag set of the first knowledge crystal and the core semantic tag set of the second knowledge crystal as semantic similarity; Respectively extracting attribute field sets of the first knowledge crystal and the second knowledge crystal; Identifying a field combination which can form a complete information chain with the attribute field set of the second knowledge crystal in the attribute field set of the first knowledge crystal; Counting the number of field combinations capable of forming a complete information chain, and taking the ratio of the number of field combinations to the total number of fields in two knowledge crystal attribute field sets as attribute complementarity; And carrying out weighted summation on the semantic similarity and the attribute complementarity, and normalizing the calculation result to a range from zero to one as the association weight between the first knowledge crystal and the second knowledge crystal.
6. The intelligent robot knowledge base management method as set forth in claim 4, wherein introducing a time dimension constraint in the initial relationship graph, and performing dynamic attenuation and enhancement operation on the edge weights, the making the relationship graph have time sequence dynamics comprises: adding a time stamp to each edge in the initial relation diagram, wherein the added time stamp records the time when two knowledge crystals connected with the edge generate association for the last time; Acquiring a time difference value between the current system time and a corresponding time stamp of each side; setting an attenuation coefficient function, wherein the input of the attenuation coefficient function is a time difference value, and the output of the attenuation coefficient function is an attenuation factor which is decreased along with the increase of the time difference value; for each edge, multiplying the original association weight with a corresponding attenuation factor to obtain a basic weight after time attenuation; monitoring newly added interaction events among the knowledge crystals, and acquiring a knowledge crystal pair related to the event when the new interaction event is detected; Searching the corresponding side of the knowledge crystal, and increasing the weight of the corresponding side by an enhancement amount based on the event intensity; and carrying out integral normalization on the weights of all edges subjected to the attenuation and enhancement operation, and updating the initial relation graph to enable the relation graph to have time sequence dynamic property.
7. The method of claim 1, wherein the performing the evaluation and the evolution tracking of the knowledge state based on the dynamic knowledge topology network specifically comprises: a state probe network connected with the dynamic knowledge topology network is established, and the state probe network is responsible for periodically collecting the activity index of the nodes and the circulation intensity of the edges in the dynamic knowledge topology network; Inputting the activity index and the circulation intensity acquired by the state probe network into an evolution analyzer; the evolution analyzer fits the overall stability curve and the local liveness distribution map of the dynamic knowledge topological network according to the historical sequence of the activity index and the circulation intensity; Comparing the overall stability curve with a local liveness distribution map with a preset knowledge health reference model; And identifying the stiff nodes and overload links in the dynamic knowledge topology network according to the comparison result.
8. The intelligent robot knowledge base management method as set forth in claim 7, wherein comparing the overall stability curve and the local liveness profile with a preset knowledge health benchmark model specifically includes: acquiring a preset knowledge health reference model, wherein the knowledge health reference model comprises a standard overall stability curve template and a standard local liveness distribution map template; Comparing the integral stability curve obtained by fitting the evolution analyzer with a standard integral stability curve template point by point, and calculating the square sum of the difference values of the two curves at each time point to be used as the integral deviation degree; Carrying out regional division ratio pair on the local liveness distribution map generated by the evolution analyzer and a standard local liveness distribution map template, and calculating cosine similarity of corresponding regional liveness values to be used as local similarity; According to the overall deviation degree and the local similarity, calculating the comprehensive difference index of the current state of the dynamic knowledge topological network and the knowledge health reference model; and comparing the comprehensive difference index with a preset threshold value, and judging the overall health state of the dynamic knowledge topology network.
9. The intelligent robotic knowledge base management method according to claim 5, wherein after identifying the stiff nodes and overload links in the dynamic knowledge topology network based on the comparison, the method further comprises: Starting a knowledge wakeup process for the identified stiff node, wherein the knowledge wakeup process comprises the steps of injecting a new association inquiry request or association context information into the stiff node; Starting a traffic grooming procedure for the identified overload link, the traffic grooming procedure comprising constructing a new auxiliary connection near the overload link or queuing a knowledge request flowing to the overload link; and feeding back the updated states of the stiff nodes processed by the knowledge wake-up flow and the overload links processed by the flow dredging flow to the dynamic knowledge topology network to complete real-time adjustment of the dynamic knowledge topology network.
10. A smart robotic knowledge base management system comprising a memory, a processor and a computer program stored in said memory and running on said processor, characterized in that said processor, when executing said computer program, implements the steps of a smart robotic knowledge base management method as claimed in any one of the preceding claims 1 to 9.

Description

Intelligent robot knowledge base management method and system Technical Field The invention relates to the technical field of intelligent robot knowledge management, in particular to an intelligent robot knowledge base management method and system. Background Knowledge base construction and management of current intelligent robots generally rely on rule-based systems, ontology engineering or static knowledge graph technology. These methods typically translate the collected data or experience into discrete facts, rules, or entity-relationship triples for storage and querying. Such knowledge representation forms are flat, and the lack of inherent logic structures and hierarchical organization among knowledge units makes it difficult for a knowledge base to effectively precipitate and express a progressive learning process of a robot from original data to high-order cognition. The accumulation of knowledge is more like a simple stack of information rather than organic growth, which restricts the deep reasoning and adaptive learning capabilities of the robot. The prior art scheme has defects in processing knowledge updating and system evolution. Once the static knowledge graph or rule base is constructed, the structure is relatively fixed. When new knowledge needs to be integrated or new association between knowledge needs to be reflected, local manual correction or global complex reconstruction is often needed, and the process is tedious and response is lagged. The knowledge system lacks an internal and active adjustment mechanism, the link relation is stiff, and dynamic reorganization and expansion cannot be performed in real time and automatically according to the change of the knowledge content and the context environment, so that the knowledge base is difficult to adapt to a complex and unsteady actual operation environment. How to effectively convert massive and heterogeneous original data generated in the operation of a robot into an advanced knowledge unit which has a stable internal structure and can be freely combined becomes a key problem. Meanwhile, how to make the knowledge units be autonomously and dynamically organized into a continuously evolving relation network, rather than a static map, so as to truly reflect the continuous growth and reconstruction process of the knowledge system along with experience accumulation is another challenge to be solved. Disclosure of Invention The invention aims to solve the defects in the prior art and provides an intelligent robot knowledge base management method and system. In order to achieve the purpose, the invention adopts the following technical scheme that the intelligent robot knowledge base management method comprises the following steps: inputting an original operation data stream generated in the operation process of the intelligent robot to a knowledge deposition module; depositing and layering the original operation data stream through the knowledge deposition module to form an initial knowledge deposition layer with a hierarchical structure; introducing the initial knowledge deposit layer into a knowledge smelting engine, and performing cross-layer fusion and recrystallization on the initial knowledge deposit layer by the knowledge smelting engine to generate a standardized knowledge crystal set; Delivering the knowledge crystal set to a topology braiding machine; carrying out space relation mapping and link braiding on knowledge crystals in the knowledge crystal set through a topology braiding machine to construct a dynamic knowledge topology network; and performing knowledge state evaluation and evolution tracking based on the dynamic knowledge topology network. As a further aspect of the present invention, the forming an initial knowledge deposition layer having a hierarchical structure by depositing and layering an original operation data stream by the knowledge deposition module specifically includes: The knowledge deposition module receives the original operation data stream and performs timeliness marking and source classification on the original operation data stream; Performing density analysis and event slicing on the classified original operation data stream, and extracting key event fragments from the data stream; depositing different key event fragments into different data levels according to semantic density and time continuity of the key event fragments; performing in-layer calibration and time stamp alignment on key event fragments deposited into each data level; Integrating all the calibrated and aligned data levels, and outputting the initial knowledge deposit layer with the hierarchical structure. As a further scheme of the invention, the knowledge smelting engine carries out cross-layer fusion and recrystallization on the initial knowledge deposit layer, and the generation of the standardized knowledge crystal set specifically comprises the following steps: The knowledge smelting engine analyzes the hierarchical structure of th