2025AD Homepage

The holy grail of autonomous driving: the double lane merge

Highways offer specific challenges for driverless cars. (Photo: Fotolia / chungking)

Article Publication Meta Data

Jerry Mooney
Jerry Mooney
Show Author Information

Article Interactions

11 shares
Rate this article on a scale of 1 to 5.
5 4 votes
12 comments
comment
0 views

Advanced driver assist technology has come a long way. To such an extent, that it is often hard to imagine why there is such a long wait in the roll out of full autonomous driving. But one especially tricky operation could slow down the pace.

The current sensors can fully shield a car in wrap-around detection with Lidar and cameras. Hi-def GPS is used to keep cars on the road as well as on the route, so what problem is left for driverless cars?

According to companies at CES 2017, including Toyota, Hyundai, Faraday future and Mobileye, the Holy Grail of autonomous driving technology is, at least in many countries, the double lane merge. Why? Because a double lane merge is not done by humans as series of rules, but an interactive series of negotiations. Drivers must attempt to merge into a lane, essentially requesting that other drivers yield. Because this creates a multitude of variables that can differ in each attempt, programming strict if/then rules into a computer program will not work.

The double lane merge, which doesn’t exist in some countries due to infrastructure and traffic rules, provides a similar problem to the fourway stop, but it is more sophisticated. The four way stop is just that, a stop – so the risk of damage is relatively low. A double lane merge is an attempt to negotiate among various cars at high speed.

When lanes merge, communication getrs tricky. (Photo: Fotolia / chungking)

This is particularly problematic among humans because some drivers are more aggressive and won’t allow a fellow commuter to enter their lane, while others are more passive. Some drivers are trying to merge from right to left while others, the opposite. This can create a conflict of goals. Ultimately, whether a driver yields or not is unknown until the attempt is made. It is precisely because of this scenario that the chief scientist at Volvo suggested that flying cars will become a reality before fully autonomous vehicles (article in German).  

Is big data the solution?

The double lane merge relies on the success of an interactive tactic: the squeeze in. The squeeze in tactic is where the driver makes it clear they want in. The other respective drivers must recognize the situation and agree to allow the merge. This doesn’t always happen, sometimes due to recognition failure, other times due to attitude or conflicts in goals. These conflicts are negotiated on the fly and are not subject to strict rules. So how do we overcome the double lane merge in order to advance autonomous driving to levels 3 and beyond?

Although the chief scientist at Volvo is skeptical and the leadership at Toyota is cautious, the consensus among car tech companies at CES 2017 was that a collaborative effort would solve this issue and others. Collaboration is required to tackle these nuanced situations, because of the enormous amounts of data necessary.

Thus, companies like Mobileye, Delphi, Intel and BMW have created partnerships, which yields more data and thus machine experience. This experience becomes ubiquitous machine wisdom as each car is networked. The wisdom is replacing programmed rules of engagement. Because fleets of cars can share their experience, they create a collective neural net that gets wiser every time any car on the network drives. And the bigger the data sets, the more reliable and wise the artificial intelligence becomes.

The fourway stop provides a problem similar to the double lane merge. (Photo: Fotolia)

With big data samples, AI can use machine learning and predictive analytics to formulate algorithms that negotiate nuanced situations like the double lane merge. The industry insiders at CES 2017 agreed that this approach is promising for solving these scenarios to within tolerable levels of reliability. However, with the fact that humans and autonomous driven cars will be on the road simultaneously, the AI must account for the human variable. A regression analysis can be used to factor each variable at the nanosecond level and decisions can be made using predictive analytics among machines.

Humans become the variable

As autonomous driving rolls out, the AI will begin to identify non-connected cars as sources of uncertainty or at least variables. Cars that share a communications network and operate from it will have known tendencies and intentions. Therefore, the human driven cars will become a potential hazard that autonomously driven cars must identify and account for and responded to with increased vigilance.

Machine learning will play a crucial role in solving the problem. (Photo: Audi)

The sentiment expressed by most auto tech companies at CES 2017 was that the momentum of the autonomous driving technology was too great to consider anything but an eventuality. So much so that even the Holy Grail of automated driving, the tricky double lane merge, doesn’t appear to be daunting anymore. These technologies must still be improved and advanced, but as of today all of the challenges appear to be solvable. So much so that much of the attention has been directed towards the experience of riding in autonomous vehicles once they become a fixture of transportation.

Here is a graphic of a double lane merge provided by Mobileye at CES 2017.

Article Interactions

11 shares
5 4 votes
Rate this article on a scale of 1 to 5.

Article Publication Meta Data

Jerry Mooney
Jerry Mooney
Show Author Information

Related Content

Quick Newsletter Registration!

Subscribe to our monthly newsletter to receive a roundup of news, blogs and more from 2025AD directly to your mailbox!