Prompt Generator for Machine Learning Plan Adaptation
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing methods for generating plans in complex environments, such as for robots and autonomous vehicles, face challenges with scalability due to large branching factors, leading to high memory and computing resource consumption.
Innovation Solution
A system that uses a prompt generator to create prompts for machine learning processes, allowing the plan generator to adapt plans based on the current environment state, utilizing large language models to break down high-level tasks into executable steps, and incorporating assertion prompts for conditional task execution.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Reliability
If plan generation searches a predefined domain in complex environments, then complete task planning can be achieved, but memory and computing resource consumption increases significantly
Solution Approach 1:
The patent segments the plan generation process into two distinct stages: (1) a machine learning model generates candidate plans from a training set, and (2) a verification process checks feasibility against environment constraints. This segmentation allows the computationally intensive candidate generation to use learned patterns while the verification step ensures reliability, reducing overall resource consumption compared to exhaustive search.
Solution Approach 2:
The system performs preliminary action by pre-training the machine learning model on a large dataset of plans before deployment. The model learns feasible plan structures and patterns in advance, enabling it to generate candidate plans efficiently during runtime without requiring exhaustive search of the entire predefined domain, thus reducing computing resources while maintaining completeness.
2Adaptability or versatility
If plan generation searches a predefined domain with many feasible actions and objects, then comprehensive task coverage is achieved, but the branching factor increases making scaling difficult
Solution Approach 1:
The patent replaces the mechanical exhaustive search system with a machine learning-based generation system. Instead of systematically exploring the predefined domain through search algorithms, the ML model learns from training data to directly generate candidate plans, substituting computational mechanics with learned patterns and reducing system complexity despite maintaining comprehensive task coverage.
3Reliability
If traditional plan generation methods are used in environments with many feasible actions, then all possible tasks can be planned, but time consumption increases
Solution Approach 1:
The patent implements feedback through the verification process that checks generated candidate plans against environment constraints and feasibility criteria. This feedback mechanism quickly identifies infeasible plans and guides the system to focus on viable options, reducing plan generation time while maintaining reliability compared to methods that explore all possibilities without such feedback.
Data Source
AI summary
Apparatuses, systems, and techniques to generate a prompt for one or more machine learning processes. In at least one embodiment, the machine learning process(es) generate(s) a plan to perform a task (identified in the prompt) that is to be performed by an agent (real world or virtual).