Making ADAS relevant in a driverless world

Recently ABI Research forecast that the ADAS market will be worth $132Bn (£90Bn) by 2026 and, by the end of 2016, analysts believe it will be worth a staggering $11bn – rising by nearly 29% compound annual growth rate.

This shows that ADAS market represents a great opportunity and potential for everyone involved in its ecosystem. However, the motivation for ADAS isn’t just about profits and earning money. From car manufacturers to technology vendors there is a genuine wish to reduce accidents, injury and death because, as the age-old adage says, “prevention is better than cure”.

James Hodgson, industry analyst, autonomous driving and location tech at ABI Research, thinks 2016 will be the year when the market moves towards combined function automation. “In 2015 it was about consolidation, bringing automatic braking and cruise control to more mass market vehicle segments,” he explains. To reduce accidents ADAS technologies need to become more widely adopted and so there will be a focus on low price and high volume vehicle segments. Examples of this include the launch of Toyota Safety Sense and Honda Sensing packages, aimed at the $300-500 price point. In late 2016 and early 2017 he thinks the next trend will begin to emerge more readily.

A combined approach

This launch will be about combining lateral and longitudinal ADAS – low level automation at highway speeds while the consolidation that began in 2015 will continue.  “There is an increasing use of sensor fusion: So rather than using just a camera or a radar, there is a move towards using a camera and radar, for example, to improve the robustness of ADAS, particularly in the case of vulnerable user protection,” he says before adding that there is a significant trend to move beyond using sensors purely for collision avoidance “to a situation in which data captured by connected, ADAS-equipped vehicles shared and aggregated for other applications.”

He adds: “Good examples include Ford’s Parking Spotter concept, and Mobileye’s Road Experience Management (REM) system for community map making. This trend brings to a close a frustrating scenario where connected cars have only used embedded sensors for collision avoidance. Alternative use cases for popular ADAS sensors, such as cameras and ultrasonic sensors will continue be very significant in future, In the meantime, empirical evidence suggest that automatic emergency braking is reducing front to rear accidents by 40%. Yet lane departure systems don’t show much of a decline in accidents because many drivers tend to feel annoyed by them and turn them off. For this reason much attention is being spent on finding an appropriate human-machine interface (HMI).

Vital infrastructure

Lance Williams, vice-president of automotive strategy at ON Semiconductor, points out that infrastructure is a key challenge, saying: “I believe that for this market of either semi-autonomous or autonomous driving, the biggest challenge is not the electronics but the infrastructure to support these systems because in the US, for example, around 30% of the roads are unpaved.” In other words they are often gravel or dirt tracks. Another issue is that reference points on the highway could be faded, which raises questions about how self-driving vehicles can learn to avoid obstacles if road markings have disappeared.

“Obstacle avoidance will be managed by the big brain, with CPUs created by companies like Mobileye, Nvidia and Intel, where you truly need to recognise what you are seeing: another car, pedestrian or bicycle,” he explains. The other approach is to look to at whether the vehicles need to avoid an obstacle or not. Questions are being raised in the market about which approach should be taken. “Subaru, for example, has been in the market for some time with a dual camera approach. Others use a single sensor approach and then you have sensor fusion where you have a combination of long range radar, short range radar, image sensing and ultra-sonic,” he adds.

Location concerns

Yet there are location-related concerns. The cars used GPS and if this becomes scrambled the driver may need to re-boot the system to allow the GPS and the dead-reckoning to work out where the vehicle is located. This functionality will be essential for autonomous vehicles to drive themselves.

In Hodgson’s view ADAS can be applied to the urban environment successfully with sensor fusion because there is a need to detect a wider array of potentially vulnerable road users, such as pedestrians and cyclists. “However, detecting these vulnerable road users is just one requirement – we also need to determine or predict their intentions,” he says. So if a ball rolls into a road and there is a child on the pavement, will the child run into the road to fetch the ball? Will it be necessary to apply the brakes or to simply drive around the obstacle? He suggested that successful collision avoidance is, therefore, as much about predicting where a potential obstacle will be, as it is accurately determining where it currently is. So, much work is being done to develop deep learning neural networks and the use of simulation enables physical testing to destruction. Yet pedestrians are still unpredictable.

Modelling and simulation

ADAS specialists and car manufacturers find that modelling and simulating activity to be an enormous challenge in situations where there is a high degree of unpredictability and randomness – particularly in an urban environment. “If you are on a highway and it’s fairly straight, then this is a simple process for a semi or autonomous vehicle but, in the urban setting, you have many more issues such as pedestrians and traffic congestion,” says Williams who’s often asked: ‘Would you like a self-driving car?’  His response is that he’d be happy to have one on a straight and long highway but he’d still like to have the ability to take control of his vehicle in, for example, an urban situation.

To overcome these challenges carmakers and organisations such as the National Highway Transportation Safety Administration in the US and Euro NCAP in Europe have come together to drive automatic emergency braking. He points out that regional differences often exist and so there needs to be some discussion about how they apply the necessary infrastructural changes and legislation. In essence the infrastructure has to become more reliable and predictable. “We have significant changes beginning to happen in infrastructure in the US but in China, for example, the roads are being changed and added almost daily,” he explains. These infrastructural changes need to be communicated in real-time because new roads will be create while some will disappear. Without this ability, autonomous driving will be impossible in his opinion as the vehicles will need to know what’s changed.

Empirical evidence

Hodgson adds: “Empirical evidence suggests ADAS is largely as effective in urban environments as it is in inter urban environments but the biggest issue to address is the penetration of systems intended for urban applications and the inclusion of pedestrian AEB in Euro NCAP ratings should encourage adoption in Western Europe.”

The evaluation of the success of ADAS in cities with respect to automatic parking, navigation systems and object recognition in order to assess the realistic timeline for mass market roll out of urban Level 3 ADAS will come from the actual use of personal and shared driving vehicles in cities, believes Williams.  “The features are going to be different for each individual or types of individual and in an urban environment you’re most likely going to be interested in self-parking and obstacle avoidance,” he says.

Market acceptance

For long distances he’s most interest in self-driving features and in navigation. He concludes that the success of ADAS and mass market roll out requires people to fully understand what the vehicle can do for them. He concludes: “The bottom line is that semi-autonomous and autonomous vehicles will prevent accidents, injuries and deaths, and in my opinion there aren’t many people who won’t like that…” That aspect they will like but, for the meantime, people will like collision avoidance technologies perhaps more than the idea of giving total driving control to the vehicle. So a mass market roll out can only be successful when the majority of drivers embrace the technology. 

Leave a comment

Your email address will not be published. Required fields are marked *