Abstract
Avoiding collisions is one of the most important issues in autonomous driving systems. The main goal
of this system is to detect objects with potential collision risk and determine if the brake has to be
triggered in order to avoid or mitigate a collision.
The two major purposes of the thesis are: first to investigate and determine the current status of the
object detection system in order to ...
[++]
Avoiding collisions is one of the most important issues in autonomous driving systems. The main goal
of this system is to detect objects with potential collision risk and determine if the brake has to be
triggered in order to avoid or mitigate a collision.
The two major purposes of the thesis are: first to investigate and determine the current status of the
object detection system in order to further develop and implement a dynamic object classification
module that can improve the emergency braking decisions when the system has to deal with
dynamic objects and second, to propose a new function called “Speed braking” with which the brake
system can be triggered on dynamic objects driving with lower speed by calculating a suitable
deceleration in each situation.
In the first part of the thesis, a description of the actual driving assistance system is presented with a
more detailed explanation of the braking system. This provides a global overview of the whole
application showing its complexity and limitations.
Second, the actual weaken cases of the brake system are presented and consequently the module
proposed to solve them. In addition, the development of the new function is explained and for both
targets the results are shown. These results reveal the dramatic improvement of the brake system.
There were 91 real traces supplied by three different clients where the expectation was not to brake
what means that the scenario was dynamic and 62 of them were wrong. With the new algorithm 60
of those 62 traces were solved, which means an improvement of 66% in dynamic scenarios. From a
general point of view, including static and dynamic situations, the solution developed represent an
improvement of 39% which represents an overall success of 99% of the traces analyzed. Moreover a
speed braking function depending on the speed of the vehicle driving in front of the host car is
implemented, working without problems in a range of speeds 2-12 km/h while driving behind a car
or a bicycle [--]
Degree
Máster Universitario en Ingeniería Industrial por la Universidad Pública de Navarra /
Nafarroako Unibertsitate Publikoko Unibertsitate Masterra Industria Ingeniaritzan
