Table of Contents

Mega space solution

For large-scale scenarios such as tourist attractions, urban districts, and large commercial complexes covering hundreds of thousands of square meters or even millions of square meters, the traditional "single acquisition, single large map" model is insufficient. This document introduces how to handle acquisition and usage challenges in mega spaces, addressing issues like excessive cumulative errors and memory overflow caused by single oversized maps.

cluste_route

Core challenges

When processing mega spaces spanning millions of square meters, attempting to create a single giant map faces multiple technical obstacles:

  1. Acquisition difficulties: Prolonged single acquisition causes device overheating and massive data volumes (hundreds of GB), making upload and processing extremely difficult.
  2. Cumulative error: Mapping errors accumulate with collection distance. Over several kilometers, even a 1% error can result in tens of meters of deviation, which is hard to correct.
  3. Performance bottlenecks: Loading and optimizing million-square-meter map data can instantly exhaust device memory.

Solution: Partitioning and fusion

The standard solution for mega spaces is partitioning. Divide the entire large scene logically into multiple physically overlapping sub-areas, acquire each sub-area separately, then fuse and optimize all sub-areas. This enables seamless roaming through dynamic loading during runtime.

Data acquisition

Follow the Mega space data acquisition method for planning.

  • Partitioning principle: Divide large scenes by area, scenery, or functional zones, with each sub-area ideally around 100,000 square meters.
  • Overlap requirement: Adjacent sub-areas require at least 200 meters of completely overlapping paths. Larger overlap areas yield better fused maps. Ensure sufficient overlap for optimizing adjacent submaps.

Before acquiring each sub-area:

Creating mapping tasks

Viewing mapping results

After mapping completes, you can:

Tip

Testing localization performance

Runtime strategy

Fused maps require no special handling—use them as single maps directly.

  • GNSS-based querying: In outdoor scenarios, GNSS assists map queries for precise positioning in large areas.
  • On-demand loading: When placing content, load each sub-area's Mesh separately.
  • Seamless switching:
    1. Localization algorithms run simultaneously in submap A and submap B.
    2. Content transitions seamlessly in overlap zones.