Project management method and system for artificial intelligence auxiliary software development

Abstract

The invention discloses a project management method and system for artificial intelligence auxiliary software development, and relates to the technical field of artificial intelligence, and the method comprises the steps: employing an artificial intelligence agent, analyzing an original demand document, outputting a conclusion, a score and a modification suggestion of document standard review, and generating a standard demand document; based on a retrieval enhancement generation technology, in combination with a text vectorization core algorithm, vectorizing the specification demand document; retrieving and analyzing similar demands in the vector database by using a similarity algorithm to obtain a demand cost report, feeding back the demand cost report to the artificial intelligence agent, and optimizing the review quality of the artificial intelligence agent by using a multi-target cost embedding algorithm in combination with code positions of similar functions; and based on the optimized artificial intelligence agent, in combination with the result code and the development duration, developing interaction data of a project related party, and analyzing and optimizing the development process of the project. According to the method, the requirement definition, the development efficiency and the overall resource utilization rate are improved.

Description

Technical Field

[0001] The present invention relates to the field of artificial intelligence technology, and in particular to a project management method and system for artificial intelligence-assisted software development. Background Art

[0002] In software development projects, ambiguous requirements often trigger a series of adverse reactions and result in wasted project resources. This stems primarily from clients' inability to understand their true needs. Secondly, requirements engineers lack the expertise to identify and analyze the driving forces behind these needs. When it comes time to design specific details during development, communication between requirements engineers and development engineers often leaves many ambiguous or ad hoc approaches unresolved. After testing and the release of a feature, clients often perceive a deviation from their expectations and only then can they formulate their desired business process. At this point, the entire feature often needs to be rebuilt and redeveloped.

[0003] However, it is common and frequent for customers to ask for the software to be launched first without being able to clearly describe their needs. This is an objective demand and cannot be regarded as the "root of all evil" in demand management. For the sake of cost control, demand engineers should come up with a relatively rough solution that is simple in results, easy to iterate, and can be launched quickly. Correspondingly, the management of this demand process should be switched to agile project management.

[0004] At the same time, interpersonal communication can sometimes lead to misunderstandings. While wasted time may have a relatively minor impact, misunderstandings often lead to inconsistent product launches, which can have a far greater impact. In project management practice, the longer the communication chain, the more likely it is to distort the original requirements. By the time requirements reach the hands of development engineers downstream in the chain, the "bullwhip effect" of the requirements becomes apparent.

[0005] In the communication chain between customers, requirements engineers, and development engineers, without considering the influence of external parties, requirements deformation is clearly caused when customers communicate with requirements engineers, and when requirements engineers communicate with development engineers. Therefore, it is important to standardize the requirements communication method and avoid any vague and ambiguous factors as much as possible. Requirement results should be accurately described on paper.

[0006] Currently, no effective solutions have been proposed for the problems in related technologies. Summary of the Invention

[0007] In response to the problems in the related art, the present invention proposes a project management method and system for artificial intelligence-assisted software development to overcome the above-mentioned technical problems existing in the existing related art.

[0008] To this end, the specific technical solutions adopted in the present invention are as follows: According to one aspect of the present invention, a project management method for artificial intelligence-assisted software development is provided, the project management method comprising the following steps: S1. Use the preset artificial intelligence agent to analyze the original requirements document, output the conclusion, score and modification suggestions of the document standard review, and generate a standard requirements document; S2. Based on the retrieval enhancement generation technology and combined with the core text vectorization algorithm, the standard requirement documents are vectorized; similar requirements in the vector database are retrieved and analyzed using the similarity algorithm to obtain the requirement cost report; S3. Provide feedback to the AI agent based on the demand cost report and use a multi-objective cost embedding algorithm to combine the code locations of similar functions to optimize the review quality of the AI agent; S4. Based on the optimized artificial intelligence agent, combined with the result code, development time, and interaction data between developers and project stakeholders, analyze and optimize the project development process.

[0009] Optionally, using a preset artificial intelligence agent to analyze the original requirements document, output the conclusion, score and modification suggestions of the document specification review, and generate a specification requirements document includes the following steps: S11. Use the preset artificial intelligence agent to perform semantic analysis on the original requirements document, identify key elements in the original requirements document, detect logical integrity, and generate conclusions, scores, and modification suggestions for the document specification review; S12. Based on the conclusions, scores and modification suggestions of the document specification review, combined with the key elements in the original requirement document, generate a structured specification requirement document.

[0010] Optionally, based on the retrieval enhancement generation technology and combined with the core text vectorization algorithm, the specification requirement document is vectorized; similar requirements in the vector database are retrieved and analyzed using a similarity algorithm to obtain a requirement cost report, which includes the following steps: S21. Based on the retrieval enhancement generation technology, retrieve the documents in the vector database, use the text vectorization core algorithm to encode the specification requirement documents into high-dimensional vectors, and store them in the vector database; S22. Retrieve historical requirements from the vector database through similarity algorithms, retrieve development time progress data, analyze and review requirement costs, and generate a requirement cost report.

[0011] Optionally, based on the retrieval enhancement generation technology, searching for documents in the vector database, using the text vectorization core algorithm, encoding the specification requirement document into a high-dimensional vector, and storing it in the vector database includes the following steps: S211. Using the core text vectorization algorithm, the specification requirement document is segmented into sentences and segments to generate initial text segment vectors. S212. Clustering and deduplication of text segment vectors are performed through similarity calculation, and vector representation is optimized to generate a high-dimensional vector. S213. Store the high-dimensional vector into a vector database and create an index.

[0012] Optionally, historical requirements in the vector database are retrieved using a similarity algorithm, and development time progress data is retrieved to analyze and review the requirements cost. Generating a requirements cost report includes the following steps: S221. Retrieve historical requirements in the vector database based on a similarity algorithm, match historical projects with semantically similar requirements, and extract corresponding progress data, including development time and resource consumption. S222. Combine historical development data with the complexity of current requirements, calculate and review the requirements costs through weighted evaluation, and generate a structured requirements cost report.

[0013] Optionally, based on the demand cost report, feedback is provided to the AI agent, and a multi-objective cost embedding algorithm is used to combine the code locations of similar functions to optimize the review quality of the AI agent, including the following steps: S31. Extract key indicators from the demand cost report, feed the key indicators as training data to the artificial intelligence agent, and optimize the parameter weights of the demand review in the artificial intelligence agent through the policy gradient method; S32. Through the multi-objective cost embedding algorithm, the semantic vectors of similar functional code locations are extracted, and combined with the key indicators in the demand cost report, the weight distribution of development cost, quality risk and reuse efficiency is dynamically balanced to optimize the review strategy of the artificial intelligence agent.

[0014] Optionally, extracting key indicators from the demand cost report, feeding the key indicators as training data to the artificial intelligence agent, and optimizing the parameter weights of the demand review in the artificial intelligence agent through the policy gradient method includes the following steps: S311. Use the key indicators in the demand cost report as reward signals, input them into the demand review of the artificial intelligence agent, and train the demand review; S312. Based on the key indicators in the cost report, the policy gradient method is used to iteratively update the parameters of the demand review in the artificial intelligence agent and optimize the allocation strategy of the demand review parameter weights; Among them, the key indicators include development cost, quality risk, and reuse efficiency.

[0015] Optionally, a multi-objective cost embedding algorithm is used to extract semantic vectors of similar functional code locations. Combined with key indicators in the demand cost report, the weight distribution of development cost, quality risk, and reuse efficiency is dynamically balanced. The optimization of the AI agent review strategy includes the following steps: S321. Based on a multi-objective cost embedding algorithm, through semantic parsing and representation learning, the semantic vectors of similar function code positions are converted into high-dimensional vectors embedded with multi-dimensional evaluation indicators; S322. Combine the key indicators in the demand cost report, perform fusion calculations on the high-dimensional vectors of multi-dimensional evaluation indicators, generate a comprehensive review strategy, and feed it back to the artificial intelligence agent for optimization.

[0016] Optionally, based on the optimized artificial intelligence agent, combined with the result code, development time, and interaction data between the developer and project stakeholders, analyzing and optimizing the project development process includes the following steps: S41. Analyze code quality, development time, and interaction data of project stakeholders through optimized artificial intelligence agents, build a multi-dimensional development efficiency evaluation model, and generate evaluation results. S42. Based on the evaluation results, optimize the development process and resource allocation strategy.

[0017] According to another aspect of the present invention, there is also provided an artificial intelligence-assisted software development project management system, which includes: a requirements specification module, a cost analysis module, a review quality optimization module, and a development process improvement module; The requirements specification module is used to analyze the original requirements document using a preset artificial intelligence agent, output the conclusions, scores and modification suggestions of the document specification review, and generate a standard requirements document; The cost analysis module is used to vectorize the standard requirement documents based on retrieval-enhanced generation technology and the core text vectorization algorithm. It uses a similarity algorithm to retrieve and analyze similar requirements in the vector database to obtain a requirement cost report. The review quality optimization module is used to provide feedback to the AI agent based on the demand cost report, and uses a multi-objective cost embedding algorithm to combine the code locations of similar functions to optimize the review quality of the AI agent; The development process improvement module is used to analyze and optimize the project development process based on the optimized artificial intelligence agent, combined with the output code, development time, and interaction data between developers and project stakeholders.

[0018] The beneficial effects of the present invention are: 1. This invention systematically optimizes the demand management and project processes in software development projects by introducing artificial intelligence technology, thereby improving the clarity of demand, development efficiency, and overall resource utilization. In response to the rework and resource waste problems caused by vague demand, distorted communication, and uncontrollable costs in traditional projects, an integrated intelligent management mechanism of "demand analysis - risk prediction - process optimization" has been established, forming a complete feedback closed loop, effectively supporting rapid response and iterative updates in an agile development environment.

[0019] 2. This invention uses an artificial intelligence agent to automatically review and standardize the original requirement documents, which not only reduces ambiguity and misinterpretation, but also lays a clear and structured document foundation for subsequent processes. At the same time, it combines vectorization technology with similar requirement retrieval to achieve a proactive assessment of development costs, which helps project parties make reasonable resource allocation and risk prevention before project establishment.

[0020] 3. The present invention realizes self-optimization of artificial intelligence agents through continuous interactive data collection and development process analysis, thereby improving the intelligence level of review and auxiliary decision-making; combining historical code, functional modules and personnel input data, it supports process optimization and quality improvement throughout the entire process, and improves the software project's ability to balance rapid delivery, low-cost investment and precise matching of needs. BRIEF DESCRIPTION OF THE DRAWINGS

[0021] In order to more clearly illustrate the embodiments of the present invention or the technical solutions in the prior art, the following briefly introduces the drawings required for use in the embodiments. Obviously, the drawings described below are only some embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying any creative work.

[0022] Figure 1 is a flow chart of a project management method for artificial intelligence-assisted software development according to an embodiment of the present invention; Figure 2 This is a principle block diagram of a project management system for artificial intelligence-assisted software development according to an embodiment of the present invention.

[0023] In the picture: 1. Requirements specification module; 2. Cost analysis module; 3. Review quality optimization module; 4. Development process improvement module. DETAILED DESCRIPTION

[0024] To further illustrate each embodiment, the present invention provides drawings, which are part of the disclosure of the present invention. They are mainly used to illustrate the embodiments and can be used in conjunction with the relevant descriptions in the specification to explain the operating principles of the embodiments. By referring to these contents, ordinary technicians in this field should be able to understand other possible implementation methods and advantages of the present invention. The components in the figures are not drawn to scale, and similar component symbols are generally used to represent similar components.

[0025] According to an embodiment of the present invention, a project management method and system for artificial intelligence-assisted software development are provided.

[0026] The present invention will now be further described with reference to the accompanying drawings and specific embodiments. Figure 1 As shown, according to one embodiment of the present invention, a project management method for artificial intelligence-assisted software development is provided, and the project management method includes the following steps: S1. Use the preset artificial intelligence agent to analyze the original requirement document, output the conclusion, score and modification suggestions of the document specification review, and generate a standard requirement document.

[0027] In one embodiment, using a preset artificial intelligence agent to analyze the original requirements document, output the conclusion, score, and modification suggestions of the document specification review, and generate a specification requirements document includes the following steps: S11. Use the preset artificial intelligence agent to perform semantic analysis on the original requirements document, identify key elements in the original requirements document, detect logical integrity, and generate conclusions, scores, and modification suggestions for the document specification review; S12. Based on the conclusions, scores and modification suggestions of the document specification review, combined with the key elements in the original requirement document, generate a structured specification requirement document.

[0028] It should be explained that the original requirement document submitted by the requirement engineer is input, and the preset artificial intelligence agent (AI Agent) conducts a preliminary analysis, outputting the document specification review conclusion, score and modification suggestions. The requirement engineer needs to sort out the standardization of the document content until it passes this stage.

[0029] The system presets the following prompt template (Prompt): As a development department manager, you now need to analyze and review the requirements specification from the following perspectives.

[0030] 1) Document completeness: whether it covers all functional, performance, and non-functional requirements (such as error handling and boundary conditions).

[0031] 2) Document correctness: the requirement description has no logical contradictions and is consistent with the user's actual scenario.

[0032] 3) Document comprehensibility: the requirements are clearly stated and unambiguous, and the development team can understand them accurately.

[0033] 4) Verifiability: Are there clear acceptance criteria and test cases?

[0034] 5) Prioritization: whether the requirements distinguish between core functions and optional functions, and whether the development sequence is reasonable.

[0035] 6) The scope of the requirements: whether it exceeds the project boundaries and poses a risk of "scope creep".

[0036] 7) Development costs, including total costs such as manpower, time, and third-party service fees.

[0037] Finally, a basic conclusion (pass / fail), score, and modification suggestions should be given.

[0038] S2. Based on the retrieval enhancement generation technology and combined with the core text vectorization algorithm, the standard requirement document is vectorized; similar requirements in the vector database are retrieved and analyzed using the similarity algorithm to obtain a requirement cost report.

[0039] In one embodiment, based on the search-enhanced generation technology and combined with the core text vectorization algorithm, the specification requirement document is vectorized; similar requirements in the vector database are retrieved and analyzed using a similarity algorithm to obtain a requirement cost report, which includes the following steps: S21. Based on the retrieval enhancement generation technology, retrieve the documents in the vector database, use the text vectorization core algorithm to encode the specification requirement documents into high-dimensional vectors, and store them in the vector database.

[0040] In one embodiment, based on the retrieval enhancement generation technology, searching for documents in a vector database, using the core text vectorization algorithm, encoding the specification requirement documents into high-dimensional vectors, and storing them in the vector database includes the following steps: S211. Using the core text vectorization algorithm, the specification requirement document is segmented into sentences and segments to generate initial text segment vectors. S212. Clustering and deduplication of text segment vectors are performed through similarity calculation, and vector representation is optimized to generate a high-dimensional vector. S213. Store the high-dimensional vector into a vector database and create an index.

[0041] S22. Retrieve historical requirements from the vector database through similarity algorithms, retrieve development time progress data, analyze and review requirement costs, and generate a requirement cost report.

[0042] In one embodiment, searching historical requirements in a vector database using a similarity algorithm, retrieving development time progress data, analyzing and reviewing requirement costs, and generating a requirement cost report include the following steps: S221. Retrieve historical requirements in the vector database based on a similarity algorithm, match historical projects with semantically similar requirements, and extract corresponding progress data, including development time and resource consumption. S222. Combine historical development data with the complexity of current requirements, calculate and review the requirements costs through weighted evaluation, and generate a structured requirements cost report.

[0043] It should be explained that the standardized requirement documents are input, the search of the internal document library is enhanced through the Retrieval Enhancement Generation technology (RAG), the requirement documents are vectorized, the similarity algorithm is used to search for similar requirements in the past, the development time progress data in git is retrieved, the team's existing historical assets and personnel skills are analyzed, and the requirement cost is further reviewed (the output of this stage is only used as a reference report for the next stage and is not mandatory).

[0044] In addition, core text vectorization algorithms (such as BERT and Sentence-BERT) map the semantic and syntactic features of requirement documents into dense vectors through pre-trained models, and similarity algorithms (such as cosine similarity) are used to quantify the semantic associations between vectors; the requirement cost report needs to integrate historical development time, resource consumption and current requirement complexity (such as function point analysis or code line estimation), and realize multi-dimensional indicator fusion through dynamic weight adjustment (such as entropy weight method or hierarchical analysis method), and finally generate a structured report containing key indicators such as development cost, quality risk, and reuse potential to provide data support for subsequent dynamic review strategies.

[0045] S3. Provide feedback to the AI agent based on the demand cost report, and use the multi-objective cost embedding algorithm to combine the code locations of similar functions to optimize the review quality of the AI agent.

[0046] In one embodiment, providing feedback to an AI agent based on a demand cost report and utilizing a multi-objective cost embedding algorithm to combine code locations of similar functions to optimize the review quality of the AI agent includes the following steps: S31. Extract key indicators from the demand cost report, feed the key indicators as training data to the artificial intelligence agent, and optimize the parameter weights of demand review in the artificial intelligence agent through the policy gradient method.

[0047] In one embodiment, extracting key indicators from a demand cost report, feeding the key indicators as training data to an artificial intelligence agent, and optimizing the parameter weights of demand review in the artificial intelligence agent using a policy gradient method include the following steps: S311. Use the key indicators in the demand cost report as reward signals, input them into the demand review of the artificial intelligence agent, and train the demand review; S312. Based on the key indicators in the cost report, the policy gradient method is used to iteratively update the parameters of the demand review in the artificial intelligence agent and optimize the allocation strategy of the demand review parameter weights; Among them, the key indicators include development cost, quality risk, and reuse efficiency.

[0048] S32. Through the multi-objective cost embedding algorithm, the semantic vectors of similar functional code locations are extracted, and combined with the key indicators in the demand cost report, the weight distribution of development cost, quality risk and reuse efficiency is dynamically balanced to optimize the review strategy of the artificial intelligence agent.

[0049] In one embodiment, a multi-objective cost embedding algorithm is used to extract semantic vectors of similar functional code locations. This is combined with key indicators in the demand cost report to dynamically balance the weight distribution of development cost, quality risk, and reuse efficiency. The optimization of the AI agent's review strategy includes the following steps: S321. Based on a multi-objective cost embedding algorithm, through semantic parsing and representation learning, the semantic vectors of similar function code positions are converted into high-dimensional vectors embedded with multi-dimensional evaluation indicators; S322. Combine the key indicators in the demand cost report, perform fusion calculations on the high-dimensional vectors of multi-dimensional evaluation indicators, generate a comprehensive review strategy, and feed it back to the artificial intelligence agent for optimization.

[0050] It should be explained that the demand cost report is input, the technical manager assigns the development engineer, the development engineer provides feedback to the AI on the problems with the demand, provides the code location of similar functions, and corrects the quality of the AI review in the previous steps.

[0051] In addition, policy gradient methods (such as PPO or REINFORCE) use key indicators such as development cost, quality risk, and reuse efficiency as reward signals to drive artificial intelligence agents to adjust parameter weights (such as convolution kernel weights or attention mechanism coefficients) during requirements review; multi-objective cost embedding algorithms (such as Transformer-based encoders or graph neural networks) analyze the contextual semantics of similar functional codes (such as function call relationships or API usage patterns), encode them into high-dimensional vectors, and combine Pareto optimization or weighted summation methods to dynamically balance the weights of cost, risk, and efficiency (for example, development cost accounts for 40%, quality risk accounts for 35%, and reuse efficiency accounts for 25%), and finally generate quantifiable and executable review strategies (such as priority sorting or defect detection threshold adjustment) to achieve multi-objective collaborative optimization.

[0052] S4. Based on the optimized artificial intelligence agent, combined with the result code, development time, and interaction data between developers and project stakeholders, analyze and optimize the project development process.

[0053] In one embodiment, analyzing and optimizing the project development process based on the optimized artificial intelligence agent, combined with the final code, development time, and interaction data between the developer and project stakeholders, includes the following steps: S41. Analyze code quality, development time, and interaction data of project stakeholders through optimized artificial intelligence agents, build a multi-dimensional development efficiency evaluation model, and generate evaluation results. S42. Based on the evaluation results, optimize the development process and resource allocation strategy.

[0054] It should be explained that the development engineer's code output, development time, and interaction data between the developer and project stakeholders should be input to analyze the efficiency, existing problems, and areas that need optimization during the development process.

[0055] In addition, multi-dimensional development efficiency evaluation models (such as regression models based on XGBoost or deep neural networks) can quantitatively analyze code quality (such as static scan defect density and test coverage), development time (such as the rationality of task decomposition and critical path time) and interaction data (such as the frequency of demand changes and communication response delays), and construct efficiency indicators (such as development efficiency index and collaboration health) through feature engineering (such as time series feature extraction or social network analysis). Optimization strategies combine reinforcement learning (such as DQN) or constrained optimization algorithms (such as genetic algorithms) to dynamically adjust agile iteration cycles, task allocation weights (such as developer skill matching) or automated tool chain configurations (such as CI / CD trigger thresholds) to achieve data-driven process improvements.

[0056] Among them, the documents submitted at the beginning and end of each stage of the project will be vectorized and stored in the vector database for future use; the first scenario is when relevant personnel actively inquire about the AI assistant, and AI uses RAG to search and answer questions; the second is to serve as a project member analysis engine. Through the interaction data of project members and the results of the responsible fields, the similarity of requirements between different projects can be compared. It can be concluded that each project member is good at the field, and the data can be used to assist decision makers in understanding project members.

[0057] According to another embodiment of the present invention, Figure 2 As shown, a project management system for artificial intelligence-assisted software development is also provided, which includes: a requirement specification module 1, a cost analysis module 2, a review quality optimization module 3 and a development process improvement module 4; Requirements specification module 1 is used to analyze the original requirements document using a preset artificial intelligence agent, output the conclusion, score and modification suggestions of the document specification review, and generate a standard requirements document; Cost analysis module 2 is used to vectorize the standard requirement documents based on retrieval enhancement generation technology and the core text vectorization algorithm. It uses a similarity algorithm to retrieve and analyze similar requirements in the vector database to obtain a requirement cost report. Review quality optimization module 3 is used to provide feedback to the AI agent based on the demand cost report, and optimize the review quality of the AI agent by using a multi-objective cost embedding algorithm and combining the code locations of similar functions; Development process improvement module 4 is used to analyze and optimize the project development process based on the optimized artificial intelligence agent, combined with the output code, development time, and interaction data between developers and project stakeholders.

[0058] In summary, with the help of the above-mentioned technical solutions of the present invention, the present invention systematically optimizes requirements management and project processes in software development projects by introducing artificial intelligence technology, thereby improving the clarity of requirements, development efficiency, and overall resource utilization. To address the rework and resource waste caused by vague requirements, distorted communication, and uncontrollable costs in traditional projects, an integrated intelligent management mechanism of "requirements analysis-risk prediction-process optimization" is established, forming a complete feedback loop that effectively supports rapid response and iterative updates in an agile development environment. The automated review and standardized output of original requirements documents by artificial intelligence agents not only reduces ambiguity and ambiguous expressions, but also lays a clear and structured document foundation for subsequent processes. At the same time, combined with vectorization technology and similar requirements retrieval, a preemptive assessment of development costs is achieved, helping project parties make reasonable resource allocation and risk pre-control before project initiation. Through continuous interactive data collection and development process analysis, the artificial intelligence agent is self-optimized, improving the intelligent level of review and decision-making support. Combined with historical code, functional modules, and personnel input data, it supports process optimization and quality improvement throughout the entire process, improving the ability of software projects to balance rapid delivery, low investment, and precise matching of requirements.

[0059] The above description is only a preferred embodiment of the present invention and is not intended to limit the present invention. Any modifications, equivalent substitutions, improvements, etc. made within the spirit and principles of the present invention should be included in the scope of protection of the present invention.

Claims

1. A project management method for artificial intelligence-assisted software development, characterized in that: The project management approach includes the following steps: S1. Use the preset artificial intelligence agent to analyze the original requirements document, output the conclusion, score and modification suggestions of the document standard review, and generate a standard requirements document; S2. Based on the retrieval enhancement generation technology and combined with the core text vectorization algorithm, the standard requirement documents are vectorized; similar requirements in the vector database are retrieved and analyzed using the similarity algorithm to obtain the requirement cost report; S3. Provide feedback to the AI agent based on the demand cost report and use a multi-objective cost embedding algorithm to combine the code locations of similar functions to optimize the review quality of the AI agent; S4. Based on the optimized artificial intelligence agent, combined with the result code, development time, and interaction data between developers and project stakeholders, analyze and optimize the project development process.

2. The project management method for artificial intelligence-assisted software development according to claim 1, characterized in that: The method of using a preset artificial intelligence agent to analyze the original requirement document, output the conclusion, score and modification suggestions of the document specification review, and generate the specification requirement document includes the following steps: S11. Use the preset artificial intelligence agent to perform semantic analysis on the original requirements document, identify key elements in the original requirements document, detect logical integrity, and generate conclusions, scores, and modification suggestions for the document specification review; S12. Based on the conclusions, scores and modification suggestions of the document specification review, combined with the key elements in the original requirement document, generate a structured specification requirement document.

3. The project management method for artificial intelligence-assisted software development according to claim 1, characterized in that: The method of vectorizing the standard requirement document based on the retrieval enhancement generation technology and combining the core text vectorization algorithm; using the similarity algorithm to retrieve and analyze similar requirements in the vector database to obtain the requirement cost report includes the following steps: S21. Based on the retrieval enhancement generation technology, retrieve the documents in the vector database, use the text vectorization core algorithm to encode the specification requirement documents into high-dimensional vectors, and store them in the vector database; S22. Retrieve historical requirements from the vector database through similarity algorithms, retrieve development time progress data, analyze and review requirement costs, and generate a requirement cost report.

4. The project management method for artificial intelligence-assisted software development according to claim 3, characterized in that: The method of retrieving documents from a vector database based on the retrieval enhancement generation technology, encoding the specification requirement documents into high-dimensional vectors using the text vectorization core algorithm, and storing the high-dimensional vectors in the vector database includes the following steps: S211. Using the core text vectorization algorithm, the specification requirement document is segmented into sentences and segments to generate initial text segment vectors. S212. Clustering and deduplication of text segment vectors are performed through similarity calculation, and vector representation is optimized to generate a high-dimensional vector. S213. Store the high-dimensional vector into a vector database and create an index.

5. The project management method for artificial intelligence-assisted software development according to claim 3, characterized in that: The process of retrieving historical requirements in the vector database using a similarity algorithm, retrieving development time progress data, analyzing and reviewing the requirements cost, and generating a requirements cost report includes the following steps: S221. Retrieve historical requirements in the vector database based on a similarity algorithm, match historical projects with semantically similar requirements, and extract corresponding progress data, including development time and resource consumption. S222. Combine historical development data with the complexity of current requirements, calculate and review the requirements costs through weighted evaluation, and generate a structured requirements cost report.

6. The project management method for artificial intelligence-assisted software development according to claim 1, characterized in that: Providing feedback to the AI agent based on the demand cost report and optimizing the review quality of the AI agent by utilizing a multi-objective cost embedding algorithm and combining the code locations of similar functions include the following steps: S31. Extract key indicators from the demand cost report, feed the key indicators as training data to the artificial intelligence agent, and optimize the parameter weights of the demand review in the artificial intelligence agent through the policy gradient method; S32. Through the multi-objective cost embedding algorithm, the semantic vectors of similar functional code locations are extracted, and combined with the key indicators in the demand cost report, the weight distribution of development cost, quality risk and reuse efficiency is dynamically balanced to optimize the review strategy of the artificial intelligence agent.

7. The project management method for artificial intelligence-assisted software development according to claim 6, characterized in that: The process of extracting key indicators from the demand cost report, feeding the key indicators as training data to the artificial intelligence agent, and optimizing the parameter weights of the demand review in the artificial intelligence agent through the policy gradient method includes the following steps: S311. Use the key indicators in the demand cost report as reward signals, input them into the demand review of the artificial intelligence agent, and train the demand review; S312. Based on the key indicators in the cost report, the policy gradient method is used to iteratively update the parameters of the demand review in the artificial intelligence agent and optimize the allocation strategy of the demand review parameter weights; Among them, the key indicators include development cost, quality risk, and reuse efficiency.

8. The project management method for artificial intelligence-assisted software development according to claim 7, characterized in that: The multi-objective cost embedding algorithm is used to extract semantic vectors of similar function code locations. In combination with key indicators in the demand cost report, the weight distribution of development cost, quality risk and reuse efficiency is dynamically balanced. The optimization of the review strategy of the artificial intelligence agent includes the following steps: S321. Based on a multi-objective cost embedding algorithm, through semantic parsing and representation learning, the semantic vectors of similar function code positions are converted into high-dimensional vectors embedded with multi-dimensional evaluation indicators; S322. Combine the key indicators in the demand cost report, perform fusion calculations on the high-dimensional vectors of multi-dimensional evaluation indicators, generate a comprehensive review strategy, and feed it back to the artificial intelligence agent for optimization.

9. The project management method for artificial intelligence-assisted software development according to claim 1, characterized in that: The process of analyzing and optimizing the project development process based on the optimized artificial intelligence agent, combined with the result code, development time, and interaction data between the developer and project stakeholders, includes the following steps: S41. Analyze code quality, development time, and interaction data of project stakeholders through optimized artificial intelligence agents, build a multi-dimensional development efficiency evaluation model, and generate evaluation results. S42. Based on the evaluation results, optimize the development process and resource allocation strategy.

10. An artificial intelligence-assisted software development project management system, used to implement the artificial intelligence-assisted software development project management method according to any one of claims 1 to 9, characterized in that: The project management system includes: requirement specification module, cost analysis module, review quality optimization module and development process improvement module; The requirement specification module is used to analyze the original requirement document using a preset artificial intelligence agent, output the conclusion, score and modification suggestions of the document specification review, and generate a standard requirement document; The cost analysis module is used to vectorize the standard requirement document based on the retrieval enhancement generation technology and the text vectorization core algorithm; use the similarity algorithm to retrieve and analyze similar requirements in the vector database to obtain a requirement cost report; The review quality optimization module is used to provide feedback to the artificial intelligence agent based on the demand cost report, and optimize the review quality of the artificial intelligence agent by using a multi-objective cost embedding algorithm and combining the code locations of similar functions; The development process improvement module is used to analyze and optimize the project development process based on the optimized artificial intelligence agent, combined with the result code, development time, and interaction data between the development and project stakeholders.

Images