Project enables self-driving cars in real-world traffic
Starting in 2017, Volvo will test autonomous driving in a large-scale trail with commuters. The vehicles to be used are latest model XC90 SUVs crammed with sensors. But the project also makes clear that sensors and sensors and algorithms in the car are not enough – many success-critical data come from outside the car.
The set of sensors in the self-driving vehicles will be sophisticated — besides a long-range 76 GHz radar in the rear-view mirror, the vehicles will have four radars behind the front and rear bumpers, one on each corner of the car. These radar sensors will detect and locate objects in all directions. By sweeping both left and right, transmitting waves that bounce off signs, poles, and tunnels, they monitor a full 360 degrees around the car.
Four cameras monitor objects in close proximity to the vehicle. Two are under the outer rear-view mirrors, one is in the rear bumper and one is in the grille. Besides detecting objects at close range, these cameras monitor lane markings. These cameras have a high dynamic range and can handle very quick changes in lightning conditions, e.g. when entering a tunnel.
A multiple beam laser scanner is placed in the front of the vehicle, below the air intake. The scanner can identify objects in front of the car and ensures very high angle resolution. It can also distinguish between objects. This laser sensor has a range of 150 metres for vehicles and covers a 140 degrees field of view.
In addition, a trifocal camera placed behind the upper part of the windscreen is three cameras in one, providing a broad 140 degrees view, a 45 degrees view and a long-range, yet narrow, 34 degrees view for improved depth perception and distant-object detection. The camera can spot suddenly appearing pedestrians and other unexpected road hazards.
Two long-range radars placed in the rear bumper of the car ensure a good rearward field of view. This technology is particularly useful when changing lanes because it can detect fast-moving vehicles approaching from far behind.
Twelve ultrasonic sensors around the car are used to identify objects close to the vehicle and support autonomous drive at low speeds. These sensors are based on the technology used for current park assist functions enhanced with advanced signal processing. A typical example of when this technology is useful is for detecting unexpected situations, such as pedestrians or hazards on the road close to the car.
All together, this represents a rather sophisticated collection of on-board sensors. It will be complemented by a high definition 3D digital map to provide the vehicle with information about the surroundings, e.g. altitude, road curvature, number of lanes, geometry of tunnels, guard rails, signs, exits, etc. The position geometry is in many cases at centimetre level.
The high performance GPS is one part of the positioning control that is enhanced by a combination of an advanced GPS, a 3-degrees of freedom accelerometer and a 3-degrees of freedom gyro. By matching the 360 degrees image created by the multitude of sensors with the map image, the car will get the information about its position in relation to the surroundings.
In addition, the vehicle complements the sensor-generated information of its surroundings by utilising V2V connectivity, and for current traffic and map data it will be connected to cloud services. "The vehicle positioning system is much more exact that standard GPS," said Eric Coelingh, Volvo Technical Specialist for Active Safety, in a web press conference. "The cloud provides the latest map data as well as current traffic data."
To guarantee the highest safety level possible, all relevant systems in the car – brake, steering system, control computers are implemented redundantly, Coelingh said. And how will the safety will be guaranteed in difficult environments, for example heavy snowfall or thick fog? "The system tests itself constantly. It has to be able to detect that its sensors cannot detect anymore. In such a case, the driver will be alerted and the automatic driving function will be disengaged", Coelingh said. "The real challenge is to make this system work not only most of the time, but also under exceptional conditions", Coelingh explained.
The trial differs from similar tests performed by other vendors in that it is the first time that a large quantity of self-driving cars will merge into the everyday traffic of a large city — albeit the test is restricted to the large traffic arteries with four or more lanes. "One of the conditions for our test was that there is a barrier between the lanes", explains Peter Mertens, Volvo’s Senior Vice President for R&D. The typical test situation will be commuting — which by the way is also the changeover scenario for automated driving at other OEMs. But for the first time, a realistic blend of real-world customers will be involved in such a trial. The driver is not obligated to stay alert and keep his view on the road; he can read mails, listen to music or even knit. "The only precondition is that he is in the driver’s seat and he is sober", Mertens said. "We do not assume that the driver can take over immediately".
How does Volvo’s strategy to introduce autonomous driving look like? "We don’t pretend that we soon have the possibility to have completely self-driving cars in cities", said Mertens. Instead, other use cases will be defined and introduced gradually, with valet parking being likely to be one of the next steps. Autonomous driving in cities will likely to be the last, the final step. "It will be an evolution", Mertens described the company’s strategy.
Related articles
Audi’s strategy to keep Google, Tesla at bay
Bosch looks to the smartphone for electromobility
Apps as driver assistant systems
Smartwatch app from BMW connects home and car
NXP extends NFC ecosystem to the car
A crash course in “the connected car”
If you enjoyed this article, you will like the following ones: don't miss them by subscribing to :
