Calibration of Perception Sensors
Typ
Examensarbete för masterexamen
Master's Thesis
Master's Thesis
Program
Production engineering (MPPEN), MSc
Publicerad
2023
Författare
Delkinov, Stefan
Reinhardt, Carl
Modellbyggare
Tidskriftstitel
ISSN
Volymtitel
Utgivare
Sammanfattning
New regulations mandating specific sensors coupled with the trend towards manufacturing trucks
with more self-driving capabilities has led to an increase in both the number and types of
perception sensors on each new model of trucks. With each new model, the need for proper
calibration becomes paramount and demands an investigation into what sensor technologies will
be required to reach a higher level of autonomy. Further, proper calibration is paramount to
ensure the accuracy concerning the physical location of the objects in the truck’s surrounding and
allowing the truck to safely navigate.
This thesis therefore explores perception sensors in the context of manufacturing trucks. The
research investigates the currently employed calibration methods used for trucks and various
calibration methods for different sensors. Both quantitative and qualitative data was gathered
from an in depth AS-IS analysis and a literature study on both calibration methods and the sensor
technologies themselves. This is done in order to form the basis for recommending concepts of
future calibration stations.
The results presented in this thesis include three variations of a high-volume factory calibration
station and one variation for a low-volume factory. The four different concepts all utilize the
same calibration methods; however, the pieces of equipment and layouts differ among them. The
concepts are capable of handling camera, LiDAR, camera stitching and Radar, with different
process times.
The conclusion is therefore that both low and high-volume factories should share the same
calibration methods enabling interoperability and scalability. The difference between them is the
amount of equipment and investment needed. For high-volume factories where time is critical
and multiple cobots are employed and utilizes parallel calibration. In contrast, low-volume
factories where time is not as critical fewer cobots are employed and utilizes sequential
calibration. This in turn entails that high-volume concepts need greater investment and are more
complex. Nonetheless both concepts can benefit from shared knowledge and resources.
The proposed concepts serve as a foundation for future calibration stations where trucks of a
higher degree of autonomy can be produced.
Beskrivning
Ämne/nyckelord
: Calibration, Sensor, Perception sensors, Radar, LiDAR, Camera, Camera Stitching, Point Cloud