27 September 2011
European project takes another step towards completely automated cars
The need to reduce road accidents and the number of people who die as a result of them has been at the forefront of automotive design for many years. But it is only more recently that electronics technology has been applied to the problem, with research teams around the world working on ways to make driving safer through providing driver assistance.
One of the latest projects to embrace the concept of driver assistance is the European HAVEit project – Highly Automated Vehicles for Intelligent Transport. Project coordinator Dr Reiner Hoeger from Continental Automotive, said the work built on statistics that showed that 95% of accidents were driver related, caused through misinterpreting information.
In Dr Hoeger's view, two possible solutions were available: "We could get rid of the driver completely, but that raised legal and technical obstacles, or we could provide driver assistance, so they are more capable of dealing with particular situations."
Dr Hoeger continued: "We wanted to remove mundane tasks from the driver, but not to overload them with information. For example, cars already have lane departure systems, which are valuable, but they only inform the driver of something. This adds to the overload; the next logical step is to allow the systems to do things for themselves."
HAVEit was not concerned with the development of new sensor technology; rather, it was more interested in using existing technology and packaging it more smartly. "We wanted to find ways to partition work between the driver and the system, but there was a lot of work to be done in making sure that people understood the system."
One of the first steps was to define vehicle assistance and to understand people's expectations of what such systems might offer. "This helped us to define different modes of operation and to make sure that drivers are still aware of what the vehicle can do," Dr Hoeger noted. "In order to avoid confusion, we developed a generic interface to keep the driver informed about which mode the vehicle is in. We then had to decide how to bundle the functionality to bring value to the driver."
Three top level objectives were: the development and validation of next generation driver assistance systems, with higher levels of automation than currently available; optimum interaction between the driver and the system, with different degrees of automated driving available; and the development and validation of a scalable and safe vehicle architecture that includes advanced redundancy management.
HAVEit has developed seven applications for cars and trucks, all aimed at enhanced safety, comfort and fuel efficiency. The applications fall into two groups: highly automated driving; and safety. Amongst the topics addressed by HAVEit were: automated queue assistance; construction site assistance; temporary autopilot; and active green driving.
One of the problems encountered by the project was the potential for drivers to rely too much on automation. "An important part of our work was to make sure the driver knows what mode the car is in, but also to make sure the driver doesn't rely too much on the system and start doing something else," Dr Hoeger noted. "So there is a driver monitoring system to avoid misuse. We can't allow drivers to say they're too tired to drive and use an automatic system."
Several strategies are available; for example, a warning telling the driver to take back control. "If the driver doesn't do this," he continued, "then other strategies are available; for example, braking so the driver knows the system is trying to communicate."
But he also acknowledged that a driver may not be able to take control. "In that case, we have to implement a minimum risk way to bring the vehicle to a defined stop, depending on the situation."
Dr Hoeger admitted the strategies still need further development. "We're not at the final stage," he said, "but we have designed the system so it can be enhanced."
The system assesses the environment and an arbitration unit decides whether the driver's wish can be executed without danger. "Another approach," Dr Hoeger continued, "is to check whether the sensing system is reliable enough and is the driver in a suitable state where automated driving can be offered."
Another HAVEit use case is assistance through roadworks. "More than 50% of people are uncomfortable here," said Dr Hoeger, "and an automated solution can help to keep traffic flowing. An experienced driver may not have a problem being 20cm away from a concrete barrier; others may think it's too close. We can help by keeping that distance automatically and by stopping the car when the car in front stops, accelerating when the traffic speeds up and so on."
A similar system is aimed at truck drivers in traffic jams. "It's boring, but the driver has to be aware," Dr Hoeger said. "This provides longitude and latitude control at speeds of up to 30km/h."
Based on information from a range of sensors, the system decides the appropriate acceleration, deceleration and steering, as well as driving speed and steering activity. The vehicle is controlled automatically, following the queue and staying in lane. The driver's alertness is monitored by an onboard camera and a warning is issued if it appears they are getting drowsy or distracted.
While the HAVEit project has now concluded, there are plans to move the developments forward. "We've proved our goals are possible with today's technology, but a lot more work still needs to be done to make it reliable enough to prove the system will always do what it says it will," Dr Hoeger noted. "We will be looking to fine tune the driver system, but will also have to do more work with people and to start discussions with stakeholders, such as insurance companies." He also pointed out that work will need to be done if the system is to be usable in urban areas.
While much of the technology used in the demonstrators has been 'off the shelf', Dr Hoeger believes challenges for the future lie in developing suitable algorithms for integrating sensor data and modelling the environment. "And we need to develop algorithms that can drive the car," he added. "During automated driving, the vehicle's actions must not surprise the driver and there is more work to be done because automated driving is 50% HMI related."
The project's work was demonstrated in a final event, held in Sweden in June. Seven vehicles were at the event, separated into two groups. Four showed driver assistance systems: automated assistance for roadworks and congestion; a temporary autopilot; automated queue assistance; and active green driving. The other three vehicles showed designs and a migration strategy for highly automated driving.
"In view of increasing traffic density, the growing flood of information available to drivers and the rising average age of the population," Dr Hoeger concluded, "highly automated vehicles will characterise the future of mobility. Automation will relieve drivers of some of the stress of driving as it guides them through traffic more efficiently, using more environmentally friendly technology."