Are there any companies currently developing intelligent software to interact with Autonomous Vehicles?
Several companies are currently working on software development for Autonomous Vehicles (AVs), including Waymo, Tesla, Cruise, and Argo. Each company continues to refine their self-driving technology to provide the safest experience possible; however, to date, no technology has been deemed fool-proof and safety continues to be of concern. A look at these technologies follows.
Waymo, Google's sister company, has quietly been developing software and technology for the AV space through its Carcraft simulation software. The company's 100 acre training center named "Castle" is located in Southern California on a former air force base, and offers the company the opportunity to run simulated tests under various weather conditions and road configurations, such as cul-de-sacs, expressways, and mountain roads. As many as 25,000 virtual cars share the road during these simulations, and are designed to increase safety features and decisions in traffic situations.
The software and simulations allow engineers to determine the decision paths that AVs will take given a number of variables. Carcraft simulations ran through 2.5 billion miles in 2016.
In November 2017, the company announced its readiness to put passengers in AVs without a backup driver, having notched 4 million driverless miles on public roadways in 23 cities around the world.
Tesla offers driverless cars with an emphasis on safety. Their software incorporates cameras that provide 360 degrees of visibility around the car, 12 ultrasonic sensors to detect objects (including other cars) around the vehicle, and software which integrates data received from the real world, such as weather and traffic conditions. Tesla continues to improve upon prior generation software versions by increasing the number of cameras and sensors, as well as improving computing power (a new onboard computer has over 40 times the power of the prior release). The result is software that can process numerous variables simultaneously and beyond human capability.
An October 2017 article addresses the slowing pace of the release of Tesla software improvements to their Autopilot system, noting the nearly five month absence of updates. Features such as rain-sensing wipers, the capability of reading speed limit signs, and functionality to differentiate between vehicle types on the road are still missing. Additional features such as expressway automated lane changes and improved steering on narrow roads were expected by the end of 2016, but have yet to be delivered. Given these missing functions, Autopilot can not yet be fully counted on, and these vehicles require some driver intervention.
An example of the partial capabilities of self-driving vehicles is the ruling by the National Transportation Safety Board (NTSB) regarding the 2016 fatal crash of a Tesla semi-self-driving vehicle. The NTSB found that the Autopilot software contributed to the crash involving a Tesla Model S sedan. While the technology is good, the author of the article (who test drove this type of vehicle) found it "abundantly clear" that Autopilot is a long way off from allowing a car to steer or drive itself.
General Motors' self-driving car unit is Cruise Automation. The company announced plans to launch a fleet of AVs in a five-square-mile section of lower Manhattan in early 2018. Each vehicle will include a human test driver who will have the capability to take control if necessary, along with one passenger in the passenger seat.
Cruise Automation already has a live app ("Cruise Anywhere") available to 10% of the company's San Francisco-based employees (roughly 20 employees). The test fleet of Chevy Bolts run 16 hours per day in the San Francisco area and had provided more than 1,000 rides through August 2017. A test driver is present for each ride to assure safety.
Cruise has had its share of vehicle accidents as well. Its AVs were involved in six crashes in California in September 2017, but most of the crashes were the result of other drivers hitting the Cruise cars as they were slowing down for stop signs or pedestrians. In one case, the driver of another vehicle was using his cell phone when he rear-ended a Cruise vehicle stopped at a red light. Thus, interacting with other drivers continues to be a challenge for AV software engineers.
The accidents did not result in serious injuries or damage.
Ford has invested $1 billion in Argo AI, a start-up developing self-driving technology exclusively for Ford's "Level 4-capable vehicles for commercial on-demand service." Level 4 vehicles are capable of taking over a vehicle's driving entirely, in certain conditions.
Over a four-year timeframe, Argo expects to develop the entire “virtual driver system,” including hardware (i.e., cameras, radar, etc.), software, and the computing platform. They will also create high-definition maps, and keep them updated over time. The goal is to create a designed exclusively for Ford’s self-driving vehicles so they can understand their current environment anywhere in the world, and know how to navigate accordingly.
Ford, too, is struggling with safety issues in their AVs. For example, a crash involving an AV occurred in Pennsylvania last year. The crash was the result of another vehicle running a red light and hitting the Ford vehicle.
Several companies are currently working on software development for Autonomous Vehicles (AVs), including Waymo, Tesla, Cruise, and Argo. Safety is a top concern for each company as they continue to refine their self-driving technologies. To date, no technology has been deemed fool-proof and safety continues to be of concern.