Örebro Studies in Technology 76 I
ÖREBRO 2017 2017DA
NIEL
R
IC
ÃO
C
AN
EL
H
AS
T
ru
nc
ated S
ig
ned D
is
ta
nce Fi
eld
s A
pp
lied T
o R
ob
oti
cs
issn 1650-8580 isbn 978-91-7529-209-0Truncated Signed Distance
Fields Applied To Robotics
DANIEL RICÃO CANELHAS
Computer Science
daniel ricÃo canelhas received his BSc in Mechanical Engineering with
emphasis on mechatronics from the polytechnical institute (IPUC) at the Ponti-fical Catholic University of Minas Gerais (PUC-MG), Brazil in 2006. Since 2012 he has been a doctoral student at the Mobile Robotics and Olfaction (MR&O) lab in the Center of Applied Autonomous Sensor Systems (AASS) in Örebro, where he also obtained his MSc in Robotics and Intelligent Systems. His research interests are in robot perception, mapping and numerical optimization under real-time constraints.
Moving around in the world and interacting with the physical objects within it are, effortless as it may seem to the average human being, rather complex tasks that require a high level of coordination between motion and perception. This thesis is about the perception part. Three-dimensional range sensor technology has improved greatly in the last decade, providing measurements at higher frame-rates and resolutions than before. More information is usually beneficial, but poses challenges in data management for robots. One map representation capable of dealing with the wealth of information produced by modern high frame-rate range sensors is the truncated signed distance field. This thesis investigates if using a truncated signed distance field as a robot’s internal representation for the outside world allows it to perform more reliably and efficiently on typical perception problems such as keeping track of where it is, knowing where it has been before and recognizing things around it.
