Transforming Plan Goal Graphs to Bayesian Networks for Autonomous Control
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Current mission planning systems face challenges in transforming Plan Goal Graph (PGG) models into Bayesian Networks (BN) due to the lack of intuitive notation and formal reasoning mechanisms, making it difficult to compute probabilistic outcomes and reason about complex mission systems, especially for autonomous vehicles.
Innovation Solution
An automated method and system that transform PGG models into BN models by reversing arc directions, adding decision nodes, and generating conditional probability tables, enabling the computation of goal achievability and plan feasibility, and providing control instructions for autonomous systems.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Ease of operation
If a Plan Goal Graph (PGG) is used to represent mission models, then the model is easy to create and understand with intuitive notation, but it lacks formal reasoning theoretical foundation and practical mechanisms to compute probabilistic outcomes
Solution Approach 1:
The patent introduces an automated transformation system that acts as an intermediary between PGG and BN. The system automatically converts PGG models into BN models, preserving the intuitive structure while adding formal probabilistic reasoning capabilities. This intermediary transformation process resolves the contradiction by allowing users to work with intuitive PGG notation while obtaining rigorous BN-based probabilistic outcomes through automated conversion.
2Reliability
If a Bayesian Network (BN) is used to represent mission models, then formal probabilistic reasoning and computation of posterior probabilities are enabled, but constructing the model directly is challenging for complex mission systems due to lack of intuitive notation for plan-goal hierarchical decomposition
Solution Approach 1:
The patent inverts the traditional modeling approach by first creating the intuitive PGG model and then automatically transforming it into the BN model. Instead of directly constructing complex BN models for mission systems, the system reverses the process: start with simple PGG notation, then use automated transformation to generate the corresponding BN structure with proper probabilistic relationships. This inversion significantly reduces modeling complexity while maintaining formal reasoning capabilities.
3Ease of operation
If PGG is used for mission representation, then plan-goal relationships are clearly separated and intuitive, but it provides no practical mechanisms to evaluate plans' feasibility and goals' achievability probabilistically
Solution Approach 1:
The patent replaces the deterministic structural representation of PGG with the probabilistic framework of BN through automated transformation. The transformation process substitutes the mechanical graph structure with a probabilistic model that enables precise measurement of plan feasibility and goal achievability. The automated system computes posterior probabilities for goals and feasibility for plans by leveraging BN's probabilistic inference mechanisms, thereby adding precise measurement capabilities while preserving the clear plan-goal relationship structure.
Data Source
AI summary
A method and system operable to perform the method is provided for control of an autonomous or unmanned system. The method includes obtaining a mission model, wherein the mission model comprises a goal and one or more assets that are used to accomplish the goal; producing, by a first hardware processor, a plan goal graph (PGG) model based on the mission model; transforming, by a second hardware processor, the PGG model into a Bayesian Network (BN) model; computing a feasibility to execute a plan and an achievability of accomplishing the goal; and providing control instructions to the one or more assets to be used to accomplish the goal.