Improving the Vision System Accuracy for Collaborative Robots
Tipus de documentProjecte Final de Màster Oficial
Data2020-07-10
Condicions d'accésAccés obert
Llevat que s'hi indiqui el contrari, els
continguts d'aquesta obra estan subjectes a la llicència de Creative Commons
:
Reconeixement-NoComercial-SenseObraDerivada 3.0 Espanya
Abstract
Robots aim to improve and facilitate the human work and their development has
extremely advanced during the past years. Year after year, the possible applications of robots
is growing and, because of that, technological advances in this industry is constant. ABB is
one of the companies that joined this industry and became a leading supplier of industrial
robots and robot software around the world. One of the latest research areas within ABB has
been towards the study of collaborative robots, meaning robots that need a high interaction
with humans both for learning tasks and applying tasks.
Consequently, this research work will be focused on collaborative robots and our aim
is to implement statistical methods to specifically improve their vision system, to understand
better how it works and facilitate human decisions. Therefore, the purpose of this project is to
generate a code which applies the Metropolis algorithm, a Markov chain Monte Carlo method,
to explore the uncertainties around the vision system’s estimations.
Two main areas will be investigated. On the one hand, in Chapter 8 we will investigate
the actual accuracy of the detected object’s estimations. For industrial applications, robots may
need to carry out work in changing environments (e.g., pick up objects even if they’re not
always in the same place). For that, a proper vision system is required. However, different
tasks may require different levels of accuracy in the vision system. Hence the importance of
this part. As a result of this investigation we have provided with a specific decision-making
proposal for different types of final uncertainty ranges accepted.
On the other hand, in Chapter 9 we will investigate the impact of the camera calibration
into the final accuracy of the calculations. Specific set-ups, depending on the number of
reference points, will be observed to analyze the impact of the camera calibration step.
To guide the reader towards these conclusions, there will first be an introduction to the
current work in robotics within ABB in Chapter 2. In Chapter 3, the reader will do a quick dive
into the theory behind computer vision, especially regarding projective geometries and camera
projections. In the same chapter, the statistical framework and the algorithm chosen will be
introduced. Then, the used programming language, Scala, is introduced in Chapter 4 and the
key parts of the code will be described in Chapter 5. Thereupon, in Chapter 6, the experimental
set-up is described and in Chapter 7 prior experiments will be performed to ensure the proper
implementation of the code and the algorithm. The two main experiments expressed before
will take place, specifying the analysis, hypothesis, and results. Finally, the impact and budget
of this project will be presented in chapters 10 and 11.
To wind up, the final conclusions are presented, and further research is encouraged.
TitulacióMÀSTER UNIVERSITARI EN ENGINYERIA INDUSTRIAL (Pla 2014)
Col·leccions
Fitxers | Descripció | Mida | Format | Visualitza |
---|---|---|---|---|
tfm-visionuncertainty-final.pdf | 2,714Mb | Visualitza/Obre | ||
appendix-visionuncertainty-final.pdf | 1,614Mb | Visualitza/Obre |