Robotic transportation: self-driving cars

Представляем вниманию читателей перевод статьи про «Автоботов» (http://controlengrussia.com/innovatsii/avtoboty/), размещенный в июльском номере Control Engineering в США.

Self-driving cars, autobots, are enabled by similar technologies as advanced industrial robots: machine vision, proximity sensors, advanced motion controls, variable frequency drives, motors, sophisticated programming, cloud connections, safety, and high-speed networks. Control Engineering Russia explained some robotic transportation advances and challenges in this version edited for Control Engineering.

01-RTEmagicC_CTLx_icon_Robot-drive-HMI_17Autonomous self-driving cars are poised to become the first widespread, high-visibility use of mobile robots. The progress of automated self-driving vehicle development in the last 10 years is astonishing. Even new automobiles without robotic capabilities look more like computers with wheels. More articles on autonomous vehicles are published daily. [And technologies also are used for mobile industrial robots and collaborative robots, such as machine vision, proximity and force sensors, advanced motion controls, variable frequency drives, motors, sophisticated programming, cloud connections, safety, and high-speed networks.]

At the dawn of the autonomous vehicles era, there were several projects proving currently used concepts and technological approaches. Several high-profile efforts were the U.S. DARPA Grand Challenge and Urban Challenge. The successor of the Junior prototype from the Stanford Racing team became the first autonomous Google Car under the direction of Sebastian Thrun. Winners of the challenge, General Motors, subsequently introduced an EN-V prototype, a tiny two-seater autonomous vehicle [and said it expects to make semi-autonomous vehicles available in 2017, along with vehicle-to-vehicle (v2v) communications. GM said it is testing fully autonomous Chevy Volts later in 2016].

Current autonomous vehicle developers have made bold statements about future robotic cars. Elon Musk, the head of Tesla Motors, had promised to have 90% cars with autonomous features by 2015. Renault-Nissan group is readying its robotic car by 2018. Mercedes has announced it would complete its truck of the future by 2025. Such short-term optimism might suggest that autonomous cars are practically here.

Elements of automated vehicles

The optimism is justified, since modern cars often include drivers’ assistants, such as global positioning system (GPS) navigators, cruise controls, anti-lock braking systems (ABS), and rear parking systems.

High-priced Nissan and Mercedes model cars are equipped with accident-prevention systems, dead-zone controls, automatic lane switching, and even autonomous driving at speeds up to 30 km/h, all steps toward fully automated vehicles. Even if a car parks itself, meets you in front of your house, or moves at snail speed by itself safely, in traffic, that’s serious progress toward fully autonomous robotic automobiles.

The number of participants in this race clearly shows that robotic vehicle technologies are rapidly developing. These include «traditional» car manufacturers (German Volkswagen, Mercedes, and BMW, Japanese Nissan, Toyota, and Subaru, and Swedish Volvo), American electric car manufacturer Tesla Motors, Google, and others, as well as start-ups. Several Russian companies are in adjacent markets. VIST Mining Technology LLC automates mine dump trucks, and RoboCV automates warehousing machinery.

Robotic retrofits

While large automobile manufacturers are developing unmanned vehicle brands, other participants are «retrofitting,» with a focus on developing navigation technologies and other technologies that turn regular cars into autonomous cars. Stakes are high, as no one wants to yield robotic vehicle supremacy to competitors. Google presented its «raw» prototype this spring, perhaps in an effort to attract cooperation with auto giants. [Google and GM had a cooperative effort in 2014.]

Interest extends beyond light vehicles. When discussing regulations for unmanned vehicles testing, California Department of Transportation included motorcycles, heavy trucks, and buses on the list. Commercial transportation holds the biggest financial interests. Regular rural routes, continuous workload, and highway driving are less complex than urban driving. Unmanned trains and automated inter-urban buses with regular routes may be easier than urban robotic taxicabs with widely varying routes and obstacles.

In October 2015, the first Formula One Russian Grand Prix was held in Sochi, and the Audi RS-7 concept car on the German Hockenheim track set a world unmanned driving speed record of 240 km/h. Car racing becomes a competition of engineering squads, and it would be very interesting to witness a race between a human racing champion and a robocar. Can track-proven technologies quickly transfer to the increased variability within big cities? That question remains.

Robotic car motivations

Why would anyone need a vehicle that travels without human guidance? Mainly, to 1) get rid of routine activities, 2) increase safety, and 3) optimize traffic flow. We still need to physically translocate ourselves and cargo over land, so we continue to spend substantial time and energy in transport. Who wouldn’t want to be productive in the car, instead of spending time behind the wheel, especially as the pace of life accelerates? Options during safe robotic door-to-door transport include reading a newspaper [or Control Engineering online or in print], finishing a presentation, safely making a business call while referencing a laptop, or napping. For now, these luxuries are for those who can afford personal drivers.

Taxi drivers, bus drivers, and truck drivers are worth mentioning. It is possible that robot vehicles may try to replace these professionals. I think a human will remain available to drive a bus or truck, to retake control or interfere in dangerous situations, as do pilots of commercial aircraft and railroad engineers [and some plant operators for some highly controlled processes]. Meanwhile, automobiles equipped with reliable autopilots will begin taking routine, perhaps more rural, routes.

On the other hand, not everyone is a good driver. The statistics of people who have been killed or injured in car crashes looks more terrifying than status reports from battlefields. When technology reaches the level needed, robotic automobiles will be much safer than inexperienced, aggressive, or drunk drivers, though some experts disagree.

A disciplined autobot as a self-driving car never goes above speed limits, keeps the correct distance, and is careful around pedestrian crossings. Some illegal parking enthusiasts or «no entry» sign violators might consider robotic drivers tedious, but autobots are already intelligent enough to break laws in ways that may reduce risk. A Google autonomous car can exceed a speed limit up to 16 km/h, adapting to the speed of traffic flow (an option introduced by a Russian computer engineer, Dmitri Dolgov). [Who would get the ticket?]

Giving increased mobility to immobile people is among the biggest advantages of future self-driving cars.

Road flow optimization is the third goal of autonomous automobile development. Many metropolitan area roads are overloaded. As some experts predict, self-driving cars will revolutionize private transportation. Just as Uber services are increasing vehicle utilization, when self-driving cars hit the market, there will be less need for everyone to own a car. Ride sharing, a lower-cost, per-use option, will decrease the number of vehicles on the roads [perhaps similar to large industrial companies purchasing functionality in service models rather than buying under-utilized assets that require maintenance].

Robotic communications

Self-driving vehicles will pre-calculate routes around traffic jams and road work, as well as communicate with nearby cars and traffic lights to optimize throughput and increase safety. Robotic cars will use a feedback approach, using elements of distributed and centralized control.

Priorities can be granted during rush hours or to certain vehicles (like an ambulance or fire crews). These concepts already are familiar in GPS navigation software that helps shorten current roadway transportation.

Technical, economic, legal

Despite the successes, technical, economic, and legal challenges remain.

While the Google self-driving car passed tests to become the first self-driving car with a certificate, supporting participative infrastructure is a problem. The test was conducted in a defined territory, so the company prepared a detailed territory map in advance. Maintenance of sustainable current road maps for widespread use of self-driving cars is still a big unsolved challenge. Google Street View is not enough for autonomous cars. Robotic vehicles need various types of information, such as road signs, road markings and traffic lights, road conditions, and roadwork, and this information needs to be updated regularly. On the other hand, autobot dependence on road signs or markings visibility has been reduced to a minimum.

Google is not the only company developing maps for self-driving cars. BMW automaker recently cooperated with Baidu search engine in an Asian market demonstration.

Secondly, the test was conducted at day, with fine weather, and good visibility. The modern technical means of navigation, such as radars, lidars, cameras, and GPS systems show unexpected performance in extreme conditions, like heavy snow or rain, fog, blinding sunlight, or very narrow streets. Another problem is the price of these sensors, and that makes self-driving cars much more expensive than vehicles with human drivers.

Most people think that autonomous cars will become viable only when there will be a majority of these cars on the road. Regular cars, unpredictable drivers, and jaywalkers add complexity. Tests show that a robotic vehicle has overcome crossroads and pedestrian crossings without problems. And traffic flow efficiency can be highly improved with v2v communication. As with industrial communications, self-driving cars should use the same language to communicate, but no standard has been set.

Self-driving car manufacturers face a question on where to test cars. Some use old military bases, some build testing areas, and others test robotic cars in real-life conditions.

The U.S. and U.K. are making progress in rules governing self-driving cars. Western governments actively support these initiatives and correspondent committees propose legislation and pass necessary laws. Several pinch points remain. Specifically, the company has to test only one self-driving car before legally claiming that all of its autonomous cars with the same system have passed this test.

An unanswered question is whose fault would it be when a car crash occurs? Is it the driver’s fault for not interfering in time, or is it the engineer’s fault because a necessary avoidance algorithm wasn’t developed; or is it the fault of a technician who made a faulty network connection? Perhaps the robot is the cause of the crash [perhaps for choosing the lower risk of two unavoidably bad options]?

What about economic barriers? The Economist has suggested that widespread use of self-driving cars may be inhibited because of negatively perceived effects on the workforce.

Some fear robots

Finally, some may avoid robotic cars out of fear, although some negative reactions toward self-driving cars may be expected. Fears may include the «rise of the machines,» loss of privacy as people may feel watched by Big Brother, or feel that they’ll become a hostage of things [locked into an automated vehicle]. While that may seem like paranoia, technology is becoming a bigger part of our daily lives, especially, perhaps, as manipulation goes beyond information to material objects.

It’s not hard to imagine that it would be easier to track and systemize movements of a self-driving car. How collected information will be used depends on who obtains the information, related laws [and cybersecurity of connected systems].

What if riders set a «shopping mall» or «restaurant» as a destination point instead of regular coordinates, and the car decides where to go? This could be convenient for riders and a sweet spot for advertisers. Maybe this is why Google is pioneering autonomous cars development.

Ссылка на сайт Control Engineering

Добавить комментарий

Ваш адрес email не будет опубликован. Обязательные поля помечены *