[mrpt-obs]
Overview
Observation classes for all kinds of robot sensors.
Library mrpt-obs
This library is part of MRPT and can be installed in Debian-based systems with:
sudo apt install libmrpt-obs-dev
Read also how to import MRPT into your CMake scripts.
Key concepts:
Sensor observations: All sensor observations share a common virtual base class (mrpt::obs::CObservation). There are classes to store laser scanners, 3D range images, monocular and stereo images, GPS data, odometry, etc. A concept very related to observations is a mrpt::obs::CSensoryFrame, a set of observations which were collected approximately at the same instant.
Rawlogs (datasets): A robotics dataset can be loaded, edited and explored by means of the class mrpt::obs::CRawlog. See also: https://www.mrpt.org/Rawlog_Format
Actions: For convenience in many Bayesian filtering algorithms, robot actions (like 2D displacement characterized by an odometry increment) can be represented by means of “actions”. See mrpt::obs::CAction.
“Simple maps”: In MRPT, a “simple map” is a set of pairs: “position”, “sensory frames” (read above). The advantage of maintaining such a “simple map” instead a metric map is that the metric maps can be rebuilt when needed with different parameters from the raw observations, which are never lost.
CARMEN logs tools: Utilities to read from CARMEN log files and load the observations there as MRPT observations. See mrpt::obs::carmen_log_parse_line and the applications: carmen2rawlog, carmen2simplemap.