Navigant Research Blog

Excitement Tempered Around Self-Driving Cars

— May 28, 2014

In recent weeks, Google has been publicly discussing the progress of testing its self-driving cars in more challenging environments.  Most of the original testing was done on California freeways, so the driving was relatively straightforward.  The more recent test routes have included a lot of local driving and urban challenges, like pedestrians and cyclists.  These scenarios have made the company more confident that its technology is getting close to being ready for production.

However, some of the original optimism has been tempered by reality.  In 2012, when the project was first made public, Google’s estimate was that the self-driving car was about 5 years away (2017).  At the time, some technology commentators and the media got very excited about the potential and were forecasting production rollouts would begin well before 2020.  In Navigant Research’s report, Autonomous Vehicles, released in the summer of 2013, we forecast that it would be 2025 before the technology would be ready for public use.  Today, Google estimates the technology will be ready in 6 years (2020).  Our forecast has not changed.

More Maps Please

It’s quite possible that Google will consider its technology ready for commercial launch in 2020, but that doesn’t mean that automotive manufacturers will have satisfied their own testing by that date.  The automakers know that everyone will be watching autonomous technology very closely, and they cannot afford any mishaps.  The testing will have to be very comprehensive.  For example, GM has been testing its own Super Cruise system for 2 years already, and the production launch was recently described as “within the next 5 years.”

But it’s not just the testing that is holding up production release.  The current Google vehicles have a rotating lidar scanner on the roof that not only costs more than the rest of the vehicle combined, but also is visually unacceptable for most, if not all, manufacturers.  The hardware development has a long way to go to be able to achieve the necessary accuracy and blend into the vehicle styling.  The Google system needs a high-resolution scanner and also relies on highly accurate digital maps.  So far, the company has developed detailed maps of lane markings, traffic signs, and signals for about 2,000 miles of road in the United States, mostly around Mountain View in California.  The United States has about 4 million miles of roads.  A lot more digitization will be required for Google’s autonomous vehicles to be used anywhere in the country, never mind all over the world.

Navigant Research believes that self-driving cars will become a reality, but that the technology will be rolled out incrementally over the next 10 to 15 years.  It’s great to see serious progress being made, but we still do not expect any major manufacturers to rush autonomous technology into production before 2025.  More details on this will come out in our updated Autonomous Vehicles report, slated for publication later this summer.

 

Autonomous Vehicles Drive Themselves toward Reality

— March 19, 2014

Australian startup Zoox made a splash at the LA Auto Show in December by hosting a stand at the Connected Car Expo to promote its ideas about autonomous vehicles.  Zoox’s view is that the best way to introduce fully autonomous driving is to start with a clean sheet of paper and develop a new type of transport from scratch, rather than incrementally changing existing vehicles.  The initial concept currently under development is a taxicab that uses a chassis made of four identical quadrants.  Each quadrant will have a wheel with its own electric motor, and all four wheels can steer.

The passenger compartment will have no steering wheel or pedal controls and will utilize a carriage layout, with passengers facing each other.  The design is being optimized for rapid prototype manufacturing techniques rather than mass production.  Zoox is targeting taxi fleets as its first customers, because the business model shows that the biggest savings come from eliminating the drivers’ wages.  The vehicles will be designed for low-speed travel on city streets.

Experience Not Required

At the Autonomous Driving conference in Berlin hosted by we.CONECT, the Zoox team actively sought feedback from the other participants.  They are in the first year of a 7-year product development plan, so there is no vehicle to sit in at present, but the overall concept is well thought out and some detail work has begun.  I am sure that the Zoox developers will be tracking progress of the Navia, a robotic driverless shuttle, and Tesla has shown what can be accomplished in the automotive industry without decades of experience.

One recurring theme in Berlin was how to develop an automated driving system that can return control to the driver safely when necessary, particularly when road conditions change beyond what the developers anticipated.  While driver assistance functions are steadily getting more sophisticated, there are huge advances to be made before people can safely be removed completely from the driving process.  Today’s incremental improvement process involves automating the control systems that have been developed over the last century for humans to use.  This seems to be the fundamental challenge that Zoox has identified, and it wants to approach the solution from the other direction.

Maps from the Cloud

In addition to the intriguing Zoox concept, the presentations at the Berlin conference were of high quality and networking opportunities were abundant.  Here are some of the highlights that I noted that gave me some fresh perspectives on the current state of autonomous driving technology:

  • Professor Emilio Frazzoli of the Massachusetts Institute of Technology pointed out that the biggest potential benefit from autonomous driving will be carsharing, far exceeding improved road safety.  His detailed analysis of traffic in Singapore indicated that 800,000 personal cars could be replaced by only 300,000 shared autonomous vehicles.
  • Dietmar Rabel, from digital map company HERE (formerly known as NAVTEQ and Nokia Location & Commerce), promoted the Internet cloud for continuous map updates and introduced the concept of crowdsourcing for accurate map data.  Rather than relying on map suppliers to continuously update the information, sensor data from connected vehicles could be shared through the cloud, thus providing near real-time updated map and road condition information locally.
  • Geoff Ballew of NVIDIA explained how his company has grown from a supplier of graphics boards for PCs into a high-performance computing specialist.  Rapid data processing will be a key requirement for self-driving vehicles to become a reality.

While the automotive industry makes slow but steady progress toward the goal of a self-driving vehicle, it’s also good to hear about new companies approaching the topic from a different perspective.  I shall continue to watch with interest as I work on an update to Navigant Research’s 2013  Autonomous Vehicles report.

 

Self-Driving Cars and Real-World Roadways

— February 16, 2014

On a recent weekend road trip, I took the opportunity to consider the practicality of an autonomous vehicle doing the driving. The 300-mile journey involved single-lane twisty country roads, dual carriageways (in U.S. terms, a four-lane divided highway), and motorways (freeways). The first part of my journey took place on a narrow country road with speed limits that ranged from 30 mph through small villages to 60 mph on the open stretches. On this route, there were very few opportunities for passing, so the driving process was relatively straightforward. A combination of the latest advanced driver assistance systems (ADAS) should be able to cope with such a drive with minimal driver input.

The next part of the journey took place on a dual carriageway, and again the driving process was quite simple, requiring that my vehicle stayed within well-marked lanes, kept to the speed limits, and avoided running into the back of slower vehicles.  All these functions could be handled by adaptive cruise control, lane keeping, and traffic sign recognition.  The one activity that would need a new system is lane changing to move to the inside lane when not overtaking. Blind spot detection would be a partial solution to this, but it would also need some highly sophisticated decision-making software.

Smoothing the Flow

The bulk of the driving took place on the U.K. motorway system, and the satnav in my car proved that it could handle giving directions to navigate the quickest route. Driving on motorways is where the benefits of autonomous vehicles would be more widespread. For some of the journey, traffic moved along briskly at the speed limit, but as vehicle volumes increased, there were periods where all lanes of traffic slowed down. If all the vehicles in the outside lane used adaptive cruise control, the traffic flow would be much smoother, and some traffic jams would be eliminated.

So the three main parts of my journey could have been handled effectively by technology that is available today. Intersections, however, represent more of a challenge. Simple traffic lights at a crossroads are not too difficult, but some roundabouts are a different matter, and will require considerable development of decision-making software. While the mechanics of driving can be replicated today, the role of the driver cannot. There are many considerations involved in driving, such as estimating closing speeds of vehicles in front and behind to decide whether it is appropriate and safe to change lanes. Anticipating what other drivers will do is another useful driving skill. It may be that an artificial intelligence system that can learn from experience will be a key component of the self-driving vehicle of the future.

10 Years Out

Some of the more advanced autonomous driving features that I outlined above will be coming to market in the next few years. As long as they are treated as driver assistance features, I believe they will be very attractive to customers and will contribute to safer and more efficient road travel. Full details about all the systems are described in Navigant Research’s recent report, Autonomous Vehicles. However, the jump to fully autonomous driving that can handle any situation remains at least a decade away. We can forget about catching up with emails or sleep while the car does the driving for many years, but the number of crashes due to driver error will surely be reduced, and soon.

One consideration for governments at present is how to encourage the development and implementation of this advanced driving technology. On one section of the trip, there was an alternative toll road to the standard highway. It appears that the majority of drivers prefer to travel on the free roads even when road work causes lane narrowing and speed limit reductions. It would improve revenue if more people used the toll road, so perhaps an incentive for drivers who use ADAS would make sense. A toll road that offered higher speed limits for vehicles with self-driving capability would both generate demand for the technology and increase road revenue.

I am looking forward to discussing these and other autonomous vehicle issues with industry colleagues at the upcoming Autonomous Driving 2014 conference in Berlin, Germany, February 27-28, 2014. I hope to share Navigant Research’s perspectives on the topic and learn more about other aspects of this rapidly evolving technology. Let me know if you will be there.

 

Making Cars Smarter – and Safer

— February 11, 2014

In early February, the U.S. National Highway Traffic Safety Administration (NHTSA) announced it will begin to promote vehicle-to-vehicle (V2V) communication technology for light vehicles.  The announcement focused mainly on the technology helping drivers avoid crashes.  Since August 2012, the Department of Transportation has been running a safety pilot program in Ann Arbor, Michigan, where about 3,000 vehicles were deployed in the largest road test of V2V technology yet.

In its early phase, the trial has centered on gathering, interpreting, and managing data from all the vehicles driving in the vicinity.  Later, the pilot vehicles will move on to sharing that information to warn against collisions.  One of the key targets is to improve traffic safety at intersections.  Other future benefits may include traffic jam and road hazard alerts.

The hardware required for V2V systems is not very expensive and most modern vehicles are already equipped with some of the necessary sensors.  V2V systems can also be easily made available as an aftermarket retrofit so that existing vehicles can participate.  The benefits of the technology, though, will remain limited until the percentage of V2V-equipped vehicles on the road is very high.

Speed Saves

Some articles in the media are quoting NHTSA as estimating that V2V could prevent up to 80% of accidents that don’t involve drunken drivers or mechanical failure.  That could be true in an ideal scenario where 100% of vehicles are operational with V2V.  But in practice, it would only take one vehicle without the communication device to cause a serious accident.  For example, when passing on a two-lane road, the V2V system could detect if a vehicle is coming the other way – but only if the oncoming car is also equipped with a working system.

Another challenge is the wireless communication technology.  To date, the V2V testing has used the 5.9 GHz frequency band.  This band has been reserved for dedicated short-range communications, a situation that is essential for safety systems because it minimizes transmission latency.  Yet, the Federal Communications Commission (FCC) is considering opening this band to unlicensed Wi-Fi devices.  In that case, it would be difficult to guarantee that V2V would react quickly enough to prevent accidents.

NHTSA is considering requiring V2V communication equipment in all new cars.  Another big challenge will be ensuring that the technology is compatible with future developments.  Requiring original equipment manufacturers (OEMs) to fit this technology is one way to kick-start the rollout of V2V, but full effectiveness will only be achieved if there are significant incentives for drivers to install the systems on existing vehicles.

 

Blog Articles

Most Recent

By Date

Tags

Clean Transportation, Electric Vehicles, Energy Storage, Policy & Regulation, Renewable Energy, Smart Energy Practice, Smart Energy Program, Smart Grid Practice, Smart Transportation Practice, Utility Innovations

By Author


{"userID":"","pageName":"Autonomous Vehicles","path":"\/tag\/autonomous-vehicles","date":"7\/31\/2014"}