Mapping and localization

Pose Engine generates pose estimates relative to a map. The map is built before localization occurs using various sensor data as described below.

What is a map?

Maps are a spatial representation of an operating site built from lidar or camera sensor data. The map is the reference against which lidar or camera localization is performed. IMUs, encoders, and GNSS/INS improve accuracy but are not strictly required for mapping.

MSA’s maps are a proprietary format that are only suitable for localization purposes; they are not suitable for annotation or storing meta information. Maps are stored as files in a directory and can be distributed using standard binary file transfer.

Maps are comprised of modular spatial chunks and are designed to support:

• Contributions from multiple robot instances from heterogeneous robot platforms
• Modular addition, subtraction, and changes to the map
• Modular distribution of specific map chunks to specific robot instances in a large managed fleet

Mapping

Mapping is the process of building a map from lidar or camera sensor data. For acausal operation, the map is built from the same sensor data used for localization.

For causal offline or online operation, mapping is a pre-localization process where one or more data collection robots drive through the area to be mapped, recording all sensor data.

Localization

Localization is the process of determining the robot’s 6DoF pose within a pre-built map using lidar or camera sensor data. Pose Engine generates pose estimates relative to the map, plus odometry if running offline in causal mode or online.

Localization does not require a GNSS or INS, but if this type of data is available, pose estimates in an earth-centered-earth-fixed (ECEF) coordinate frame can also be generated.

Map support and quality signals

Pose Engine generates map support and localization quality metrics alongside each pose estimate. These signals allow downstream systems to condition their behavior on current and projected future pose reliability.

Map durability

Pose Engine maps are designed for long-term reliability in dynamic environments. Maps remain usable in the presence of:

• Third-party movers: people, vehicles, and other robots in the environment
• Seasonal changes: lighting, vegetation, and environmental variation over time

Regions that change significantly over time can be re-mapped and updated without rebuilding the full site map.