Giving robots the ability to operate in the real world has been, and continues to be, one of the most difficult tasks in AI research. Since 1987, researchers at Carnegie Mellon University have been investigating one such task. Their research has been focused on using adaptive, vision-based systems to increase the driving performance of the Navlab line of on-road mobile robots. This research has led to the development of a neural network system that can learn to drive on many road types simply by watching a human teacher. This article describes the evolution of this system from a research project in machine learning to a robust driving system capable of executing tactical driving maneuvers such as lane changing and intersection navigation.
In this article we introduce Boss, the autonomous vehicle that won the challenge. Boss is a complex artificially intelligent software system embodied in a 2007 Chevy Tahoe. To navigate safely, the vehicle builds a model of the world around it in real time. This model is used to generate safe routes and motion plans both on roads and in unstructured zones. An essential part of Boss's success stems from its ability to safely handle both abnormal situations and system glitches.
Before autonomous trucks and taxis hit the road, manufacturers will need to solve problems far more complex than collision avoidance and navigation (see "10 Breakthrough Technologies 2017: Self-Driving Trucks"). These vehicles will have to anticipate and defend against a full spectrum of malicious attackers wielding both traditional cyberattacks and a new generation of attacks based on so-called adversarial machine learning (see "AI Fight Club Could Help Save Us from a Future of Super-Smart Cyberattacks"). As consensus grows that autonomous vehicles are just a few years away from being deployed in cities as robotic taxis, and on highways to ease the mind-numbing boredom of long-haul trucking, this risk of attack has been largely missing from the breathless coverage. It reminds me of numerous articles promoting e-mail in the early 1990s, before the newfound world of electronic communications was awash in unwanted spam. Back then, the promise of machine learning was seen as a solution to the world's spam problems.
Uber and Volvo announced an agreement where Uber will buy, in time, up to 24,000 specially built Volvo XC90s which will run Uber's self-driving software and, presumably, offer rides to Uber customers. While the rides are some time away, people have made note of this for several reasons. I'm not clear who originally said it -- I first heard it from Marc Andreesen -- but "the truest form of a partnership is called a purchase order." In spite of the scores of announced partnerships and joint ventures announced to get PR in the robocar space, this is a big deal, but it's a sign of the sort of deal car makers have been afraid of. Volvo will be primarily a contract manufacturer here, and Uber will own the special sauce that makes the vehicle work, and it will own the customer.
Waymo's autonomous cars have steadily rolled through test routes in multiple states over the past few years, and now the company claims it has passed a new milestone: 4 million self-driven miles logged on public roads. That makes the Waymo fleet the most experienced autonomous car platform currently on the road, according to the company, which says the average American driver would take 300 years to hit the same mark. While the number is arbitrary to a degree, the progress it represents is essential to Waymo's mission to create truly driverless cars. The AI behind the platform needs to be trained in real-world situations to understand how to react to every potential condition it might face, so the more test miles it logs, the better. The Google spinoff says its fleet of test vehicles drove the last million miles in just six months, a rapid improvement from the 18 months it took to accumulate the first million (from the first public test).
"The Tesla's automation did not detect, nor was it required [to], nor was it designed to detect the crossing vehicle," Robert L. Sumwalt, chairman of the National Transportation Safety Board, said at the start of a hearing reviewing the Florida crash. Tests by the National Highway Traffic Safety Administration determined that Tesla and other vehicles with semiautonomous driving technology had great difficulty sensing cross traffic. The NTSB staff also said that Tesla's reliance on sensing a driver's hands on the wheel was not an effective way of monitoring whether the driver was paying attention. The NTSB staff recommended the use of a more effective technology to determine whether a driver is paying attention, such as a camera tracking the driver's eyes.
The University of Michigan opened the $6.5m, 32 acres Mcity, the world's first controlled environment specifically designed to test the potential of connected and automated vehicle technologies that will lead the way to mass-market driverless cars Ford has become the first major car maker test autonomous vehicles at Mcity – the full-scale simulated real-world urban environment at the University of Michigan. Occupying 32 acres at the University's North Campus Research Complex, M City includes approximately five lane-miles of roads with intersections, traffic signs and signals, sidewalks, benches, simulated buildings, street lights and obstacles Occupying 32 acres at the University's North Campus Research Complex, it includes approximately five lane-miles of roads with intersections, traffic signs and signals, sidewalks, benches, simulated buildings, street lights, and obstacles such as construction barriers. For the most part, self-driving cars will be ready for the open road long before the open road is ready for them. That's true for the private companies designing and building self-driving cars, and for the taxpayer-funded government agencies that design and build the roads on which they'll drive.
Lyft is adding yet another new name to its growing list of official self-driving partners -- and this one already has an established track record of putting autonomous cars on real life streets. The ride hailing company just announced a new agreement with nuTonomy, the MIT-founded startup that was the first to test a self-driving program carrying passengers alongside real traffic on city streets. The new partnership between Lyft and nuTonomy will aim to put self-driving cars on the streets of nuTonomy's home city, Boston, where the startup has been conducting road tests since late last year. Lyft now has three deals tied to autonomous car development in place; the ride hailing company also has pacts with Google's Waymo and GM, which also invested $500 million in the company and holds a seat on Lyft's board of directors.
Hurricane Matthew's record rains were but the first of many obstacles faced by millions of evacuees in Florida, Georgia, and the Carolinas this past week. Roads were blocked by chest-high floodwaters and downed trees. Gas stations ran out of fuel. And traffic sat backed up for miles along interstate highways as floodwaters overtook what appeared to be tens of thousands of households. Most did make it to safety, thanks to evacuation orders, well-planned emergency procedures, and traffic managers switching up lanes to move a glut of vehicles (contraflow for the win).
This means that programmers must account for every type of road situation a car may encounter. MIT's Technology Review spoke with Amnon Shashua, CTO and cofounder of the technology firm to learn more about the initiative. Mobileye has been in the news of late for another reason--its system was the one being used by the Tesla vehicle that was involved in a car crash in Florida recently--the incident is still under investigation by the NHTSA. Tesla publicly blamed Mobileye, and because of that, a rift developed between the companies, which are now no longer partners. Shashua does not believe that will harm the company's new initiative, though--building a system based on neural networking, which, if all goes according to plan, will allow a car or truck to learn how to drive in much the same way that humans do.