Page 66 - Data Science Algorithms in a Week

50 Olmer Garcia and Cesar Diaz

in which the number of such deaths is relatively high. Figure 1 shows historical data for
traffic accident deaths in Brazil, USA, Iran, France, and Germany. However, the per
capita statistics are controversial as the number of people who drive varies between
countries, as does the number of kilometers traveled by drivers. There is a significant
difference in the statistics between developing and high-income countries.

Figure 1. Traffic accident deaths per 10,000 citizens. Sources: Brazil (DATASUS), United States
(NHTSA), Iran Bahadorimonfared et al. (2013), Germany(destatis.de) and France (www.securite-
routiere.gov.fr).
The trend toward the use of automated, semi-autonomous, and autonomous systems
to assist drivers has received an impetus from major technological advances as indicated
by recent studies of accident rates. On the other hand, the challenges posed by
autonomous and semi-autonomous navigation have motivated researchers from different
groups to undertake investigations in this area. One of the most important issues when
designing an autonomous vehicle is safety and security (Park et al., 2010). Currently,
machine learning (ML) algorithms have been used at all levels of automation for
automated vehicles (NHTSA, 2013):

 No-Automation (Level 0): The driver has complete control of the vehicle, but
machine learning helps through the perception of the environment to inspect and
alarm the driver.
 Function-specific (Level 1) and combined Automation (Level 2): One or more
primary driver functions – brake, steering, throttle, and motive power – are
controlled in the specific moment for the algorithms, like lane centering
algorithms or adaptive cruise control. In these systems, the conventional