Concept

Map representation

You might say: map type

A map representation is the data structure used to store the world for localization, mapping, and planning.

What it does

Sets which spatial questions the planner can ask and how much memory and computation the map needs.

Ask

Does the planner need occupancy, landmarks, raw geometry, or a surface?

What to look for

1. Occupation probabilities assigned to cells
2. Distinctive landmarks across observations
3. Registered 3D samples from scans
4. Smooth surfaces from a depth volume

Catalogue: Map

• Occupancy grid: An occupancy grid discretizes the environment into cells and assigns a probability of occupation to each cell.
• Landmark / feature map: A landmark map stores positions of distinctive features and their associations across observations.
• Point cloud: A point-cloud map stores registered 3D samples from one or more scans.
• Voxel / TSDF / mesh: A TSDF integrates depth into a voxel volume and can generate smooth surfaces.

Sources

1. Cesar Cadena et al., Past, Present, and Future of Simultaneous Localization and Mapping: Toward the Robust-Perception Age, Map representationsThe former models the environment as a sparse set of landmarks, the latter discretizes the environment in cells and assigns a probability of occupation to each cell.
2. Open3D, TSDF Integration Tutorial, TSDF integration and surface extractionTSDF integration reduces noise and generates smooth surfaces.
3. Timothy D. Barfoot, State Estimation for Robotics, Chs. 8–9It is an example of simultaneous localization and mapping (SLAM) since we are going to estimate the robot and landmark positions.

Use it on your project

teardown-apply: hold "Map representation" (map-representation) from the teardown at https://teardownatlas.com/robotics/ against my project

Paste it into a coding agent that has the teardown skills.

An AI agent wrote this map with the teardown skill. It can contain errors: check the sources before you rely on a part. About and legal