Future of ADAS

 

Autonomous driving is hailed as ‘the technology’. There is more content than one can read in the lifetime about life changing potential of autonomous driving technology. As per Market Watch: The Advanced Driver Assistance Systems (ADAS) market was valued at $19.8 billion in 2021 and is anticipated to grow at a CAGR of 21.4% all along the forecast period.

While ADAS is a linear approach towards autonomous driving, we are still in transition from the second stage of ADAS to the third. There are quite a few constraints which hold back this transition, and we will come to them in a minute. That being said, there are disruptive solutions coming up from tech companies which aim at addressing level 4 autonomy challenges.

Where is this all heading? How far are we, in a true sense, from our dream of self-driving cars? While projecting a tight timeline is next to impossible, here are some constraints which either need de-bottlenecking. As you will discover in the article, some of them are already being eased out.

 

The Trolley Problem

When a machine is set to replace humans, challenges are not just technical, they are also ethical. The  popular ‘Trolley Problem’ discusses ‘choice’ of selecting between two groups of pedestrians in case of an inevitable accident. Generally, humans do such choices while they drive but when it comes to embedding such choices in an algorithm, is it ethical? While trolley problem is very simple on paper, it gets pretty complex on field as the situations become more nuanced. A large number of such ethical dilemmas need resolution to speed up our journey towards next levels of ADAS.

 

Digital Radars

Major reason causing failure of present ADAS/ autonomous driving systems is dependence on legacy, analog radars. Analog radars offer poor performance and do a meagre job at detecting pedestrians, especially children in low low visibility situation. Digital radars offer much better resolution and contrast (16 times and 30 times respectively as claimed by some manufacturers) as against analog radars. So far, poor radar performance has been a big (or narrow!) bottleneck. Easing it out through digital radar will accelerate our journey towards the next levels of autonomy.

 

Sensors                                               

As we ladder up the ADAS levels, the number of interactions between vehicle and driver, vehicle and surrounding and vehicle and infrastructure becomes increasingly complex and the number of sensors required increases exponentially.

For instance, a fully autonomous vehicles is expected to have 25-30 cameras and number of radars and LiDAR’s. Whereas an ADAS level 2 vehicle as lesser than 8 cameras, at max 2 radars and in very few cases, a single LiDAR. This highlights how dependent ADAS upgrades are on availability of auto grade sensors.

The commonly available industrial sensors cannot be used to automotive applications as their performance is lackluster. Even sensors used in airplanes won’t work satisfactorily as they are designed to work when the pilots are present in the cockpit. Developing high performance sensors that too at a price point which will make their use at huge volumes feasible is something we have not yet achieved. This is a major wall that needs to be demolished.

 

Computing Power

Another area where a lot of work is happening and needs to happen is the processors and computational excellence. As we just saw in the previous paragraph, as ADAS levels advance, the number of cameras, sensors, radars and LiDAR’s goes up exponentially. Well, data being generated by these components needs to be processed in real time so that decisions can be made real time. Developing powerful computers that can process huge amounts of data at lightening speeds is a tough nut to crack. Another challenge is figuring out a power source for these power-hungry computers.  Terms such as sensor fusion are actually at the stage of infancy, and we need them to mature real quick to move towards the next levels of ADAS.

 

So, what’s the path ahead?

 

The road towards full autonomy is hard one. But we also see a lot of dedicated efforts being put by OEMs, start-ups and enthusiasts. Almost all the pain points addressed above are already identified and being tackled by numerous players. We will definitely see breakthroughs happening much faster in the near future.