Closing The Gaps To Realize Full Autonomous Driving

Autonomous driving will be fully embraced when the public perceives that automobiles are as smart as humans in recognizing and responding to unusual circumstances.

But human drivers are imperfect and unpredictable. So, it is easy to imagine that the potential number of unusual circumstances that can be encountered on the road are infinite and impossible to envisage.

The ability to distinguish between true threats from false alarms on the road is at the crux of solving the problem that currently hinders driver assistance (ADAS) and autonomous driving technology. Identifying, assessing, and responding to real-world driving challenges in a robust, fail-safe and feasible manner is what will ensure a safe road ahead for drivers, pedestrians, and other vulnerable road users. So, the question becomes, can all of the technology required to put an autonomous vehicle on the road address this gap? Can we engineer the safe and effective handling of outlying road circumstances and secure the acceptance of the general public? For the industry, it is time to advance from theoretic possibilities to actual implementation for automakers across the globe.

At this point, typical radar does very well at detecting moving objects around the vehicle and potential hazards. Radar is also more equipped to deal with real road conditions that vehicles will inevitably encounter because it can also detect objects in all environment conditions – snow or sleet or in bright lights or dark areas (where cameras cannot detect objects). However, the infinite number of potential corner cases on the road for AVs demands that the sensors in the vehicle develop higher levels of detection so that true safety can be achieved.

A new innovative sensor - 4D Imaging Radar that is an ultra-high-resolution solution, closes the gap by providing two orders of magnitude higher resolution than current radars. It can support 100,000 detections per frame. It provides unique functionalities such as detection of hundreds of objects, separation of objects by elevation, elimination of false alarms, and detection of stationary objects.

The detection by a 4D Imaging Radar is so effective because it is done in ultra-high resolution in both azimuth and elevation in a long range and a wide field of view so that nothing can go unnoticed. In the past, the AV industry has addressed what is lacking by simply tacking on additional components within the vehicle’s sensor suite; the “more is merrier” approach. But, in reality, high sensitivity, high-resolution 4D Imaging Radar, when coupled with a camera, serves as a minimalist, cost-effective approach for advanced capabilities and offers unique advantages to perception algorithms that are developed by OEMs and Tier 1s.

4D Imaging Radar plus camera can be utilized as a standalone perception unit within the sensor suite. And, when used in the right way, this combination can address relevant road applications and provide high level perception abilities that senses the vehicle’s ego-velocity, localizes the vehicle in its lane, and provides precise and accurate free space mapping (an accurate map of the available space around the vehicle) to distinguish drivable from non-drivable environments. The ability to detect pedestrians, bicyclists, scooters and sense stationary objects would have helped prevent the cause of the recent ADAS accidents.

These accidents were highly publicized bumps in the road towards the goal of full autonomy—a zero-road fatality reality. But ADAS features, such as forward collision warning, blind-spot monitoring systems, and lane departure warning, are continuing to be developed. As of May of 2018, according to AAA, at least one ADAS feature can be found in 92.7 percent of vehicles available in the U.S. and the financial commitment to include more than one ADAS feature is now within the reach of many new car buyers. The technology is popular, of course, because it goes a long way to reduce the amount of accidents, serious injuries, and fatalities resulting from car collisions. As ADAS technology improves in the future, so will the safety benefits. This has always been the promise as we move closer to fully autonomous vehicles.

But we are closer than ever to achieving this promise. Radar is being revolutionized with capabilities that make it look and perform like an entirely different sensor elevating radar to one of the most integral components of the sensor suite. And 4D Imaging Radar’s promise for safety is no longer just a vision. It is an actuality that is happening now and one that we will see on the ADAS and AV road as soon as 2022.

Kobi Marenko is CEO of Arbe