GAT Docs

GAT Platform

GAT Documentation

GAT is a cloud-edge tri-model framework that fuses learning, analytical constraints and logical control for verifiable robotic task planning.

Overview

This page is the documentation shell for the GAT World site. The final library will be populated from Markdown files supplied by the backend, while this route already uses a real documentation layout: persistent section navigation, article typography, source-oriented examples and an on-page table of contents.

Markdown backend pendingWhen the source corpus is ready, each document can map into this layout without changing the page structure.

Tri-model loop

GAT documentation should explain the system as a task-completion framework. The model loop combines learned scene understanding, analytical feasibility checks and logical control rules before execution is sent to the robot runtime.

LayerResponsibilityOutput
Learning modelInterprets scene state, task intent and candidate operations.Task sequence proposal
Analytical modelChecks spatial, temporal and physical constraints.Validated plan state
Logic controllerChooses execution order, replanning boundaries and failure handling.Executable task policy

Cloud-edge runtime

The docs should separate cloud-side planning from edge-side execution. Cloud services can maintain task context, run heavier reasoning and aggregate simulation evidence; the edge runtime should preserve collision constraints, local sensing and low-latency robot control.

  • Cloud planning: mission decomposition, simulation rollouts and global consistency checks.
  • Edge execution: local perception, collision-safe actuation and task-state feedback.
  • Closed-loop update: execution evidence returns to the planner for refinement.

Markdown source

The eventual documentation library can be driven by front matter and article bodies. The page is ready for a generated navigation index, article routes and Markdown rendering.

---
title: Tri-model Framework
section: GAT Platform
order: 2
---

# Tri-model Framework

Explain the learning model, analytical model and logic controller as one task-completion pipeline.

Next steps

Once the backend provides Markdown, the placeholder sections should become real article routes with search, versioned navigation and source-backed diagrams. The current page keeps the documentation surface clean until that content lands.