Transition Probability Model for Plan Generation
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing plan generation technologies, such as those described in WO 2018/220744, often fail to generate optimal plans under exceptional cases due to insufficient reflection of information from exceptional situations, leading to suboptimal results.
Innovation Solution
A computer system that calculates and modifies transition probabilities between processes based on historical data, using a model management system to evaluate and adjust probabilities for reliability, ensuring accurate plan generation even under unusual conditions.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Reliability
If a model is generated based on past plans using conventional methods, then the plan generation works well for common cases, but it fails to generate optimal plans under exceptional cases due to insufficient reflection of information from exceptional situations
Solution Approach 1:
The system performs preliminary actions by generating multiple candidate plans including alternative process sequences before evaluating them. When exceptional cases are detected during evaluation, the system has already prepared alternative plans that can be selected, rather than generating new alternatives on-the-fly. This preliminary generation of multiple possibilities ensures that exceptional cases are covered in advance.
Solution Approach 2:
The plan generation system dynamically adapts its behavior based on the evaluation results. When exceptional cases are detected during plan evaluation, the system can switch to selecting from pre-generated alternative plans or adjust the generation process. This dynamic response allows the system to handle both common and exceptional cases effectively, transitioning between different operational modes as needed.
2Productivity
If only common case information is reflected in the model, then the model is simple and efficient, but it cannot handle exceptional cases appropriately
Solution Approach 1:
The system generates a larger number of candidate plans than strictly necessary for common cases, including alternative process sequences that may be needed for exceptional situations. This excessive action of generating more candidates ensures that when exceptional cases occur, suitable plans are already available among the candidates, maintaining both efficiency and reliability.
Solution Approach 2:
The system changes parameters during the plan generation and evaluation process. When exceptional cases are detected during evaluation, the system can modify generation parameters to focus on alternative plans or adjust the selection criteria. This parameter adaptation allows the system to maintain efficiency for common cases while ensuring reliability when exceptional conditions arise.
Data Source
AI summary
A computer system holds model management information for managing a model for calculating, based on a feature amount of a process pair generated based on a plan history and formed of two processes, a transition probability of the two processes forming the process pair. The computer system comprises: a transition probability calculating unit uses, in a case of receiving input data including a plurality of target processes, the model management information and a feature amount of a process pair formed of a reference target process and a transition destination target process, to thereby calculate a transition probability of the process pair; and a transition probability modifying unit modifies the transition probability of the process pair determined to be unreliable. The transition probability calculating unit generates a new plan by determining the order of execution of the plurality of target processes based on the transition probability of the process pair.