Feature-Based Robotic Mapping with Generalized Labeled Multi-Bernoulli Filters for Planetary Landers


The feature-based robotic mapping (FBRM) problem is concerned with building, maintaining, and refining a feature map, such as terrain, pieces of furniture in a room, or obstacles in a parking lot, about a mobile sensor whose trajectory is assumed known. The nature of the problem necessitates that an FBRM method account for a possibly varying number of map elements, spurious measurements from the sensor, and missed detections. Filtering methods that employ random finite sets are generating substantial interest among the research community, and recently, approaches that append labels to the individual set elements have shown impressive theoretical support and performance. This work aims to illustrate the use of one of these tools, the generalized labeled multi-Bernoulli filter, for FBRM and compare it to some alternative random finite set-based filters. The method is demonstrated in a simulation where a lunar lander traverses an uncertain terrain as a camera on-board the vehicle takes measurements of the feature locations as pixel-coordinates in the camera's image plane. These observations are used to feed the corrector stage of the filter to perform FBRM as the vehicle descends along its trajectory.

Meeting Name

AIAA/AAS Astrodynamics Specialist Conference (2016: Sep. 13-16, Long Beach, CA)


Mechanical and Aerospace Engineering

Keywords and Phrases

Astrophysics; Bandpass filters; Cameras; Mapping; Robotics; Set theory; Signal filtering and prediction; Feature location; Filtering method; Measurements of; Missed detections; Random finite sets; Research communities; Robotic mapping; Spurious measurements; Planetary landers

International Standard Book Number (ISBN)


Document Type

Article - Conference proceedings

Document Version


File Type





© 2016 American Institute of Aeronautics and Astronautics (AIAA), All rights reserved.

Publication Date

01 Sep 2016

This document is currently not available here.