IEEE Consumer Electronics Magazine - March/April 2021 - 19

Figure 1. Visualization of common scenarios in
ACC, FCW, LKA, and BW.

more and more vehicles with these features
become available. Figure 1 shows an overview of
the four features we focus on for Level 2 autonomy, which are discussed in the following.
ACC was first introduced in the Mercedes-Benz
S-Class sedan in 1999, with the goal of increased
driver comfort. An ACC causes the ego vehicle to
follow a lead vehicle at a specified distance (see
Figure 1) without exceeding the speed limit specified by the operator upon activation of the feature.10 If the lead vehicle slows down, then it is the
responsibility of ACC to slow down the ego vehicle
to maintain the specified distance. Although
implementations differ, all ACC systems take over
longitudinal control from the driver (see Figure 1).
The challenge in ACC is to maintain an accurate
track of the lead vehicle with a forward facing sensor and using longitudinal control to maintain the
specified distance while maintaining driver comfort (e.g., avoiding sudden velocity changes).
An LKA is an evolution of lane departure
warning systems. It involves a forward-facing
sensor (often a camera) to identify where the
lane lines exist in front of the ego vehicle. Once
the lane lines have been detected (e.g., using
Canny edge detection and Hough transforms on
forward-facing images), LKA can then determine
if the ego vehicle lies between those lines (see
Figure 1). If the ego vehicle appears to be drifting
toward a position where it will cross lane line
boundaries, LKA engages steering torque to
steer the vehicle in the opposite direction of
the lane line until it no longer has the trajectory
to cross that lane. The LKA systems have been
known to over-compensate, creating a " pingpong " effect where the vehicle oscillates back
and forth between the lane lines.11 The main
challenges in LKA are to reduce this ping-pong

March/April 2021

effect and the accurate detection of lane lines on
obscured (e.g., dirt covered) roads.
An FCW uses information gathered via various
forward facing sensors to determine whether the
ego vehicle is going to collide with an object in
front of it (see Figure 1). As objects approach the
boundary where the vehicle can no longer come
to a stop, an audio-visual warning notifies drivers
instructing them to apply the brakes. As this is a
safety-critical system, it is important that FCW
avoids false positives as well as false negatives to
improve driver comfort, safety, and reduce rearend accidents.12 For this to be achieved, it is a
necessary prerequisite that the sensors used by
the FCW system be placed where they have an
accurate view of the vehicle in front of them. The
United States National Transportation Safety
Board has recommended that FCW be included
in all new vehicles.13
Finally, BW uses sensors mounted on the
sides of the ego vehicle to determine whether
there is a vehicle toward the rear on either side
of the ego vehicle in a location the driver cannot
see with their side mirrors14 (see Figure 1). This
area is typically referred to as the " blind spot "
and must be verified as clear of any vehicles
before the driver can attempt to make a lane
change. Without BW, the driver must turn their
head to make that verification on their own.
With BW, the driver can maintain their concentration on the road ahead. As BW requires information about a specific area near the rear of the
vehicle, it is a challenge to find an optimal sensor
placement that maximizes the view of the blind
spot. If the sensor is too far forward, it will miss
the blind spots entirely, causing a vehicle accident when the driver makes a lane change. If the
sensor is too far back, it will end up capturing
information for areas around the ego vehicle
that are not in the blind spot, decreasing the
sensor's effectiveness at viewing the presence of
vehicles surrounding the blind spot.

Feature Performance Metrics
To quantify the performance of a sensor
configuration on a vehicle being evaluated over
drive cycle test cases (i.e., across various driving
scenarios; see " EXPERIMENTS " ), we define eight
metrics (m1-m8) that are characteristic of the
configuration's ability to track and detect

IEEE Consumer Electronics Magazine - March/April 2021

Table of Contents for the Digital Edition of IEEE Consumer Electronics Magazine - March/April 2021