A self-driving vehicle developed by researchers in Germany has undertaken a monster 2,400 km (1,500 mi) journey from the US/Mexico border at Nogales to Mexico City without any guidance from a human hand. While the lengthy road trip took place mostly on highways, the AutoNOMOS car also had to contend with potholes and city streets before safely pulling into Mexico City to complete the longest trip ever completed by an autonomous vehicle in Latin America.
The AutoNOMOS car is a collaborative research project involving scientists from Germany's Freie Universität and the University of Navada, Reno. The vehicle has been approved for use in Berlin since 2011, where it has already taken to freeways and city streets. It has also been tested on roads in the US and Switzerland, but traversing the Mexican landscape would present a new set of challenges for the researchers.
The team spent one year readying the car for its epic journey, which involved adjusting its driving parameters to account for things like potholes. Last month, in cooperation with the University of Nevada, Reno, some members of the research team collected data from 6,000 km (3,730 mi) of freeways driving through parts of the US and Mexico to produce the detailed navigation maps required for the car.
Fitted with seven laser scanners, nine video cameras, radar at the front, back and sides and GPS, the AutoNOMOS car gathers information about its surroundings, monitoring the positions of nearby vehicles, pedestrians and traffic lights. A host computer then uses this information to control the car in accordance with traffic laws.
The AutoNOMOS' route would take it from the US-Mexico border across four Mexican states, through tropical regions, the semi-arid Sonoaran Desert and over mountain ranges. The roads varied from new freeways to old narrow roads without markings, passing through construction sites and urban zones along the way.
"We started at Nogales," says Professor Raul Rojas from the University of Nevada, who also holds a joint appointment with Freie Universität. "We covered 250 to 300 miles (400 to 480 km) daily, so it took a week to arrive to Mexico City. Some parts of the highway were scary, but we had no important safety incidents. The Federal Highway 15 in Mexico goes through a few big cities, such as Guadalajara. A significant issue is the absence of lane markings in long segments of the highway that have been just repaved after damaging Pacific thunderstorms over the summer."
Ideally, the self-driving car accelerates and brakes with subtlety and avoids unnecessary steering. The researchers described the results of their first Latin American jaunt in this regard as "very impressive," with the car largely achieving these aims and smoothly passing other cars even at speeds of 130 km/h (80 mph). The car is also reported to have responded appropriately to dangers on the freeway.
The Freie Universität team will look to build on its success in Mexico by getting started on its next autonomous vehicle next month. The scientists are working on technology that would allow the car to discern the intentions of other road users, along with making the system cheaper and more compact.
"We now want to design a miniaturized driving system, one in which the sensors and computer are no longer visible, and that would also be much more affordable," says Daniel Göhring, from Freie's Department of Mathematics and Computer Science.
You can see AutoNOMOS hit the road in Mexico in the video below.
Source: Freie Universität
I note that it uses GPS. Seeing as GPS signals can be jammed, one of the major hurdles that they need to overcome is how should the system cope with such an incident? Imagine the chaos that would be wrought in the rush-hour if all the vehicles within a mile or so radius suddenly had no idea where they where and where they needed to go. Amplify the jamming signal and whole cities could be taken out. And what happens when the jamming ceases? As a very minimum the car systems must negotiate with each other, otherwise it could get very nasty (search for 'road rage' on YouTube!).
If it doesn't already exist, there needs to be a steering committee (no pun intended) that can act as a forum for all interested parties to express their needs and concerns and to determine areas of standardisation so that any autonomous vehicle system can listen to any sensor, talk to any actuator and read a standardised map that includes absolutely all the information that any autonomous vehicle system might need, no matter how unlikely. Provision of this map and how it is managed (especially how it is updated, both centrally and in-vehicle has to be key to the success, or otherwise, of this whole autonomous vehicle experiment.
Even detailed maps down to the meter are not enough and could never be live updated on the fly due to massive data transmission requirements over vast areas. Nothing ever remains the same, potholes, stuff falling onto the road, crossing, or falling onto the roadway, flash floods, etc.
Yes machines can react to things faster than humans, but the old saying garbage in garbage out still stands. Unless a vehicle can see, hear, feel without fail everything around them all the time in any condition, than driverless is fruitless...