Ensuring Advanced Driver Assistance Systems Make Reliable Predictions

Advanced driver assistance systems (ADASs) are found almost everywhere and, in some cases, are mandatory for new vehicles in developed markets. Such systems typically include sensors that measure the proximity to other road users, objects and hazards including pedestrians, cyclists, and vehicles. These proximity measurements need to be accurate within a few percent to make appropriate control decisions.

In a white paper commissioned by Light, a depth perception technology company, Guidehouse Insights outlines the requirements for accurate, high resolution depth measurement in the perception systems of ADASs and automated driving systems (ADSs).

According to the white paper, great strides have been made in improving occupant protection in crashes over the last 50 years. However, there is still work to be done, as pedestrians and cyclists have experienced the greatest increase in fatalities during the last decade.

“Whether a vehicle is equipped with ADAS or a full ADS, the ability to accurately measure the position and trajectory of other road users, particularly vulnerable road users, is essential to bringing down these numbers,” said Sam Abuelsamid, principal research analyst with Guidehouse Insights. “Achieving that goal will require getting crash avoidance technologies into as many vehicles as possible, which in turn means the technologies must be easy to integrate into vehicles and affordable to purchase.”

The white paper provides an overview of active sensors such as radar and lidar, along with passive sensors, including cameras of various types used in these systems. It also discusses newer multi-camera solutions with wide-spaced sensors that can bridge the gap between costly active sensors and less accurate traditional cameras by providing the ability to take accurate, physics-based distance measurements.

“Because cameras are widely regarded as a core element of almost every ADAS and ADS, every automaker or supplier should consider how to use a physics-based multiview solution as part of the software stack,” Abuelsamid said.