Augmenting CILEMP to Enable Fleet Autonomy with Generative AI (CILEMP-FLAG)

A Generative AI augmentation to the KnexPlan planning suite that extracts complex planning models and situational data from structured text and generates Navy-compliant operations orders (OPORDs), reports, and C2 messages in natural language.

Agency

Office of Naval Research (ONR)

Capabilities

AI Research & Innovation

AI Applications

01 / Customer Challenge

The challenge our customer brought us.

Existing automated planning systems rely on structured data to encode environment information, possible actions, and other relevant planning information.

They cannot effectively ingest operational knowledge or extract critical plan models and situation data from unstructured sources like text reports or human subject matter expert (SME) inputs. This leads to three key limitations: development requires slow, expensive model engineering using specialized languages; operators require extensive technical skills to work with automated planning systems; and semantic interoperability between diverse autonomous platforms is challenging, hindering the scale of fleet autonomy.

02 / Knexus Approach

What we designed and delivered.

CILEMP-FLAG uses a novel Generative AI architecture to automate the conversion between human interpretable descriptions and structured representations that can be used by automated planning systems:

Planning Knowledge Extractions: Ability to convert between natural language and structured planning representations even when those representations are complex and have limited examples/documentation. Provides the ability for non-technical users to modify planning knowledge, decreasing the skill requirements and time required.

Support for Complex Representations: Use of an agentic-inspired approach to subdivide the extraction process into simpler subproblems, allowing CILEMP-FLAG to generate planning knowledge even when more complex representations are used (e.g., that include timing, scheduling, and continuous numeric values).

Plan Queries: Provides the ability for a user to query system outputs (e.g., plans, simulation logs) and get clear, natural language responses on topics of interest, performance, and other relevant topics.

03 / Mission Impact

The difference it made.

CILEMP-FLAG streamlines and scales autonomous planning systems by removing the need for specialized model engineering languages, allowing developers to generate models from plain text.

This allows advanced planning technology to be used even when access to technical experts is limited, allowing for rapid changes to planning knowledge, assumptions, and capabilities. In addition to the increased warfighter readiness and strategic planning capabilities, CILEMP-FLAG also provides significant manpower, time, and cost savings.

More Case Studies