| KEYWORD |
A simulator implementing NVIDIA Kepler GPUs to evaluate the reliability of self-driving cars
keywords GPU, PEDESTRIAN RECOGNITION, DEPENDABILITY, SAFETY
Reference persons STEFANO DI CARLO
External reference persons Alessandro Vallero (alessandro.vallero@polito.it)
Research Groups TESTGROUP - TESTGROUP
Thesis type RESEARCH
Description Recent years have witnessed an increasing attention to self-driving cars. Self-driving cars employ a wide number of techniques to sense the environment. Computer vision is one of the most promising, allowing the car to “visually” inspect the surrounding by means of a camera. The acquired images are processed by GPUs to classify and detect objects. Reliability of GPUs and computer vision algorithms occupy a key role in this scenario, influencing strongly the safety of the passengers, the pedestrians and the surrounding environment. The study of the reliability of self driving cars is a hot research topic from an industrial and an academic perspective.
This thesis proposes the study of a micro-architecture simulator for NVIDIA Kepler GPUs to evaluate their reliability and the reliability of the computer vision algorithm employed by self-driving cars. This work is a collaboration among three different universities: the Federal University of Rio Grande do Sul, the University of Athens and the Politecnico di Torino.
See also kepler_isa.pdf
Required skills - strong skills in programming languages (C/C++, Python, Bash);
- basic knowledge of computer architecture and/or GPU architecture;
- experience with Linux environment;
Deadline 16/04/2019
PROPONI LA TUA CANDIDATURA