Robotic Task Anticipation Using LLMs and Classical Planning
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing robotic task anticipation methods rely heavily on data-driven deep networks and LLMs, which struggle with multistep, multilevel decision-making, require extensive training, and fail to leverage domain knowledge, leading to inefficiencies and difficulty in adapting to environmental changes.
Innovation Solution
Integrate large language models (LLMs) with classical planning using a standardized prompt generation and clustering approach, combined with Planning Domain Definition Language (PDDL) to predict and optimize task anticipation, enabling proactive task planning and efficient execution.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Reliability
If data-driven deep network architectures and LLMs are used for task anticipation, then task prediction capability is improved, but training time and processing power requirements increase significantly
Solution Approach 1:
The patent pre-processes and structures task descriptions into standardized formats with defined schemas, predicates, and relationships before training. This preliminary organization of data allows the model to learn more efficiently from structured information rather than raw unstructured text, reducing training time while maintaining prediction accuracy.
Solution Approach 2:
The patent transforms task anticipation from a pure learning problem into a structured reasoning problem by changing the parameter representation from raw text to formal logical structures. This parameter transformation enables the system to leverage domain knowledge and reduce reliance on extensive training data and computational resources.
2Extent of automation
If LLMs are used to generate task plans, then automation level is improved, but ability to leverage domain knowledge and adapt to environmental changes deteriorates
Solution Approach 1:
The patent introduces an intermediary layer between the LLM and the task execution system. This intermediary translates high-level LLM-generated plans into formal PDDL representations that can be processed by classical planners. This intermediary layer enables the system to combine the generative capabilities of LLMs with the rigorous reasoning and domain knowledge integration of classical planning methods, improving both automation and adaptability.
Solution Approach 2:
The patent segments the task planning process into distinct stages: high-level task generation by LLM, formal representation in PDDL, classical planning for detailed execution, and adaptation to environmental changes. This segmentation allows each component to specialize in its strength while working together as an integrated system, maintaining automation while improving adaptability through modular design.
3Adaptability or versatility
If classical planning methods with PDDL are used, then domain knowledge integration is improved, but complexity of implementation increases
Solution Approach 1:
The patent creates a universal framework that can handle multiple task domains and complexities through a single standardized PDDL interface. The system uses general-purpose PDDL templates and schemas that can be adapted to different domains without requiring domain-specific planning algorithms, reducing implementation complexity while maintaining strong domain knowledge integration capabilities.
Data Source
Figure 1
Figure 2
Figure 3A~3B
AI summary
The present invention generally relates to the field of robotics, and, more particularly, to a method and system for task anticipation by integrating large language models and classical planning. Conventional methods for task anticipating use data-driven deep network architectures and Large Language Models (LLMs) for task estimation but they do so at the level of high-level tasks and require a large number of training examples. Thus, embodiments of present disclosure provide a method and system for task anticipation by integrating large language models and classical planning. The disclosed method and system leverages the generic knowledge of LLMs through a small number of prompts to perform high-level task anticipation, using the anticipated tasks as joint goals in a classical planning system to compute a sequence of finer granularity actions that jointly achieve these goals.