Testing and Modeling of Visual Perception Systems for Intelligent Vehicle Applications


Staff Mentor:

prof. J. Hellendoorn


Other Mentor(s):

Dr. Arturo Tejada Ruiz (TNO)

Keywords:

Intelligent control; Machine learning; Optics and imaging

Description:

Introduction

The Integrated Vehicle Safety (IVS) department at TNO develops technology for automated driving and cooperative mobility for intelligent vehicle applications. An important aspect in all such applications (truck platooning, autonomous cars, etc.) is the ability of the vehicle to recognize and represent its environment. This is generally accomplished through an array of onboard sensors, such as cameras or LIDARS, combined with advanced data processing algorithms.

The current challenge is to correctly characterize the properties of such visual perception systems (in terms of precision, robustness, etc.) under different use (highway, city, platooning) and maneuvering (lane keeping, lane changing) conditions.

Assignment description

The main goal of this assignment is to design and implement a test methodology for visual perception systems that allows one to compare the output of such algorithms to measured ground truth data. Moreover, the outcome of the project should be a model relating the visual perception system parameters (frame rate, vehicle speed, etc.) to the accuracy of its output.

Main Elements

• Design and implementation of a test setup for (at least) the TNO camera-based visual perception systems

• The tests should take into account highway driving and both lane keeping and changing.

• The test may require field work (for ground truth collection).

Added value

For the project:

• Robust characterization of our current visual perception systems

• Standard test methodology for future visual perception systems

For the intern:

• Know how on visual perception systems for intelligent vehicle applications

• Hands on experience with state of the art automated road vehicles

• Working in a dynamic environment on automated driving functions

© Copyright Delft Center for Systems and Control, Delft University of Technology, 2017.