Autonomous long distance trains announced for Australia

Mining giant Rio Tinto will invest U$518 million in autonomous trains for its long distance heavy haul rail network. The company plans to put the first autonomous train into operation in 2014. Rio Tinto currently operates 41 trains from its Australian mines to ports with 148 locomotives and 9400 iron ore cars.

The company expects productivity improvements because of greater flexibility in train scheduling and the removal of driver changeover times. Besides increased network capacity, they also expect more efficient fuel use and thus lower carbon emissions.

Generally trains are much better suited for autonomous operation than road-based vehicles because of their fixed tracks. Unfortunately, very few truly autonomous driverless trains are in operation today. While some cities have driverless commuter systems, these typically operate in carefully controlled environments where most of the intelligence is located within the rail network and little intelligence within the locomotive itself. The Rio Tinto approach needs to be different: because of the size of the rail network (1500km) most of the intelligence will have to be placed within the locomotive. Hopefully Rio Tinto will be able to demonstrate quickly that significant productivity improvements are possible by using autonomous trains and thus start the transition towards more efficient and cost effective public transportation systems. It remains to be seen, however, to what degree labour unions and train regulators will be able to limit progress in this area.

Image © Copyright 2012 Rio Tinto

Green license plates for driverless cars

Nevada’s Department of Motor Vehicles has now issued the regulations for operating autonomous cars. In testing mode, they will carry red license plates and they must have at least two occupants. One of them must be able to assume control over the vehicle at all times. Once an autonomous car has been certified for fully autonomous operation, the car will display green license plates (Nevada’s  standard license plates are blue-gray).

Key issues of the regulation:

  • Whether physically present or not, the operator of an autonomous car shall be deemed the driver with respect to applicable traffic laws.
  • Operators of autonomous cars require a special Nevada driver’s license with a ‘G’ amendment. The amendment can be easily obtained for the holder of a Nevada drivers license and costs $5.
  • Autonomous cars may be sold in the state once they have passed testing and certification by the car manufacturer or an autonomous technology certification facility.
  • Autonomous technology certification facilities will be private entities and will be licensed by the Department of Motor Vehicle.

The regulation requires the following components within each autonomous car:

  • A black box which captures the last 30 seconds and retains its data for 3 years
  • A visual indicator when in autonomous mode inside the vehicle
  • An easily accessible, non detracting mode switch for engaging or disengaging autonomous mode
  • A failure routine which allows the car to transfer control back to the operator when it detects an error or which safely moves the car out of traffic and brings it to a stop should the operator not be able to assume immediate control.
  • It does not explicitly mention a mechanism to limit autonomous mode to the geographical area (e.g. Nevada highways) where autonomous mode is legal.

The regulation contains provision for the testing phase

  • Minimum 1 Mio U$ deposit or surety bond (maximum 3 Mio U$ for more than 10 vehicles)
  • Report any accident the autonomous car is involved with within 10 days to the Department.

This is a landmark regulation, setting up precedent on which other laws will be modeled. The regulation does not really solve the responsibility problem by deeming the operator the driver. It is clearly modeled on the assumption that the autonomous vehicle will be privately owned by Nevada residents, and does not consider the issue of car-sharing and operating autonomous vehicles as taxis (Would the taxi’s passenger be considered the operator and therefore need a Nevada driver’s license with ‘G’ amendment?).

By addressing the case where the operator is not present within the car, the regulation already looks ahead to the time where autonomous cars will not be limited to highways but may roam all public roads!

 

 

Florida’s autonomous car bill much weaker than Nevada’s

The United States is quickly establishing itself as the leader in autonomous car technology. Several states are currently introducing bills addressing autonomous vehicles. Last summer Nevada became the first state to allow the operation of autonomous vehicles on state highways as long as conditions to be spelled out by the Department for Transportation are met.
Currently two bills adressing autonomous cars are moving through Florida’s house (HB 1207) and senate (SB 1768). However these bills are not in the same league as the Nevada law (AB 511):

  • While the Nevada law explicitly allows both the operation and testing of autonomous vehicles, the Florida bill only considers ‘the purpose of testing the technology‘.
  • In Nevada, autonomous vehicles need not be under the control of a human whereas the Florida bill specifies that the ‘operation of the test vehicle must be continuously monitored in a manner that allows active control over the vehicle to be immediately assumed by a human operator‘.
  • The Florida bill requires the Department of Transportation and Motor Vehicles to draft a report recommending legislative action related to autonomous vehicles by February 1, 2014. In contrast, the Nevada bill authorizes its Department of Transportation to ‘adopt regulations authorizing the operation of autonomous vehicles‘ and specifies the various issues which need to be addressed.
  • The Florida bill only has one area where it is wider in scope than the Nevada law: It applies to all public roads whereas the Nevada bill is limited to state highways. Of course, as a human needs to be able to assume immediate control at all times in Florida, there is little difference to the status quo anyway. Google has already driven driverless cars thousands of miles under such terms on public roads in California.

In summary, little seems to be gained from the Florida bill. It does not contain reasons for autonomous car manufacturers to head to this state and will not help to establish Florida as a leader in this nascent industry.

 

Thrun to teach free online course on programming robotic cars

Driverless car pioneer and Stanford professor Sebastian Thrun will share his knowledge in this 7 week interactive online course. Thrun is not only the first winner of Darpa ‘s 2005 Grand Challenge competition for autonomous vehicles. He is also employed by Google where he leads their mostly secret Google cars project.

This will be a serious course, with a university-level work load, assignments and exams. Last fall Thrun and Peter Norvig taught their joint introductory class on Artificial Intelligence at Stanford in two versions: One version to their Stanford students and an online version with exactly the same content to approximately 160000 students from 190 countries. More than 23000 students completed the course!

After this experience Thrun co-founded Udacity, a company that believes that  “university-level education can be both high quality and low cost”. Through this venture Thrun simultaneously advances education and the adoption of autonomous cars. This course will excite thousands of students for this topic, advance the mind-share for autonomous cars and may even send future employees towards the Google robocars team.

The course CS373 is free, has few requirements – knowledge of the Python programming language would be helpful – and starts on February 20.

BMW shows highly automated driving on the Autobahn

BMW’s research group has released a video of a modified BMW 5 Sedan driving autonomously on the German autobahn. The car is equipped with lidar, radar, ultrasound and camara sensing technologies. The car drives completely autonomously on the autobahn. It switches lanes automatically, recognizes speed limits, and can cope with a wide range of unusual situations. As the project’s leader Nico Kämpchen explains this requires exact knowledge of highway lanes – with centimeter precision! All the sensors have been integrated into the car in an unobtrosive way to avoid spoiling the car’s good looks. The car has now logged almost 5000 km of autonomous driving on the autobahn. While this is a research project, we expect that such technology will appear in high end models in the not so distant future.

Posted in BMW

Some of the legal issues raised at SCU law symposium

Last week’s symposium on the legal issues of autonomous vehicles addressed many interesting issues. While the full range of the discussion will only be available to the public with the upcoming special issue of the Santa Clara Law Review (expected in May), here are some of the more interesting issues:

  • If a driverless car drives a drunk person home, could the person be charged with drunk driving? This is by no means a trivial issue because all current autonomous cars allow switching between driverless and normal modes. In an accident doubts may be raised about who had control of the car. A similar problem arises for operating mobile phones while driving. The Nevada law on driverless cars includes an explicit provision for this case.
  • The law changes introduced in Nevada and being considered in Florida and Hawai only address the legality of operating driverless cars in traffic. They don’t affect the key issue of liability.
  • Will driverless cars be able to recognize a police officer and be able to understand their directions? How about roadside workers?
  • How about privacy concerns about the travel data that’s collected by onboard computers
  • Potentials for misuse because of hacking, obtaining remote control over the vehicle and failure because of high dependency on global positioning infrastructure.

Frank Douma, a transportation expert at the University of Minnesota, concluded: “There’s probably as much work to be done on regulations and cultural acceptance as there is on refining the technical or engineering obstacles to these things”

Conference to explore the legal implications of driverless cars

Autonomous car technology is maturing fast but legal issues may soon become the biggest obstacle to their adoption. This emerging topic will take center stage at a symposium to be held in Santa Clara, CA on January 20. Organized by the Santa Clara Law Review, the day long event brings together researchers and officials from the National Highway Traffic Safety Administration. Issues to be addressed include liability aspects, insurance issues and implications for the regulatory framework. More information is available here.

NPR interview: Where is driverless car technology now?

National Public Radio featured an interview about the state of driverless car technology on January 17. Bryant Walker Smith, a legal scholar at Stanford estimated that driverless car technology could be available on the market within 10 years. He discussed progress on the legalization of driverless cars in Nevada and Florida and stressed the great potential to save lives. A transcript and soundtrack of the interview is available here.

 

 

Google awarded driverless car patent

Google is building a patent portfolio in driverless car technology. In December, the US Patent Office awarded Patent 8078349 for transitioning a car from human-driven to autonomous mode. A key problem which this patent solves is to ensure that the car knows its precise location when switching into autonomous mode. GPS may not be precise enough for the vehicle to understand where it is (it may only be accurate by 10 meters) and to determine the direction it should take. Therefore Google proposes using markings called landing strips – which may be embedded in the road. The human driver stops the car on a landing strip which the car then detects and uses to determine its location. Landing strips may even contain an embedded QR-code. But landing strips need not be marked on the ground. They may consist of recognizable well defined locations which a car can detect by examining its surroundings and for which it can look up the data in a database or online. An example could be a conventional, clearly marked parking spot.

The patent also addresses additional issues of providing instructions to the autonomous car when switching to autonomous mode. This could include instructions to move to a different location where the car is needed or to proceed to a service station for maintenance.

Futuristic autonomous vehicle design Autonomo 2030

Australian designer Charles Rattray has let imagination run wild with his Autonomo 2030 concept car pictured below. He envisions a fully autonomous electric car with a multi-layer nano-scale surface which includes solar cells. Windows can be switched between opaque and fully-transparent modes. Multiple such cars can automatically travel together in distances of 20 cm thus significantly reducing drag and energy consumption. The autonomo is not much longer but significantly narrower than the Mini and can carry up to two adults seated behind each other. Linked in a flock, two Autonomo vehicles should be able to use a single lane on many streets. Combined with inter-vehicle communication and synchronization this should do much to eliminate traffic jams.For more info see the detailed concept.

Autonomo 2030 Concept Car

 Image source, license