“Self-driving vehicles to revolutionise our mobility by 2030”
16 Jun - 2017
The
expectations concerning automated driving are running high. Some 95
percent of the technology is ready but ‘the last percentages are the
hardest’, Professor Dariu Gavrila is to state on Friday 23 June in his
inaugural address at TU Delft. Most of the scientific challenges lie in
the complexities of urban traffic and in dealing with cyclists and
pedestrians. How to recognise intent of other road users and how to
drive in a socially acceptable way in human-inhabited environments will
be the key research problems that TU Delft’s new Intelligent Vehicles
group will address.
‘Every
year, about 1.2 million people die in traffic accidents worldwide. More
than 90 per cent of these deaths are attributable to driver error.
Introducing self-driving vehicles can prevent these errors,’ Professor
Dariu Gavrila of the TU Delft 3mE faculty will claim in his inaugural
address. ‘In addition, you will be able to make better use of the time
you currently spend driving your car – being productive, communicating,
or indeed just relaxing. Sharing driver-less cars could also result in
greater traffic efficiency, with fewer vehicles. This means that the
space which is now used for parking could be redesignated in order to
make towns and cities more liveable.’
Last few percentage points The technology for self-driving
vehicles is 95 percent ready, says Professor Gavrila, who spent twenty
years working in the R&D division of Daimler, the maker of
Mercedes-Benz. ‘As is so often the case, those last few percentage
points are the hardest to cover. From a scientific perspective, the
highway scenario is largely resolved – the first self-driving cars (with
limited functionality) are already on the market. However, urban
traffic is a different matter. That scenario is much more complex
because of the crossing traffic, unclear street markings, the numerous
traffic signs and traffic lights, and the proximity of other road users
such as cyclists and pedestrians.’
Interaction and anticipation An area where present-day
self-driving vehicles have difficulty is their interaction with other –
conventional – road users. Vehicles have to be able to anticipate – to
recognise the intentions of others in time in order to be able to
respond accordingly. Otherwise, vehicles will have to operate
excessively cautiously. ‘In principle, you could allow a self-driving
car to move in Amsterdam without a driver, but it would have to drive at
5 km/h and stop for everything,’ says Gavrila. ‘That would not be a
driving style that is socially acceptable.’
Machine learning techniques using
big data can figure out what road users look like and how they
generally behave, and thus can help to better assess traffic situations.
Conversely, driverless cars would have to indicate their intentions to
those around them in an intuitive way. Furthermore, self-driving
vehicles will have to operate in a way that is comfortable for their
occupants. ‘These are precisely the issues that our new Intelligent
Vehicles group at TU Delft is working on,’ explains Gavrila.
View from inside the Intelligent Vehicle demonstrator of TU Delft - Intention recognition of other road users is key
Market introduction There is no simple answer to the
question “when will we see self-driving cars?” First of all, it depends
on the interpretation of ‘self-driving’. For example, will people have
the role of ‘back-up’ driver in the event of emergencies? How versatile
is the area in which self-driving functionality is being offered? Under
what conditions will the system work (weather, day/night)? What driving
conditions will be possible (speed, free or only behind a lead vehicle)
and what support will be needed from the infrastructure? As well as
technological issues, there are still unresolved matters regarding
amendments to and harmonisation of legislation. ‘Ultimately, it will be
customer acceptance and trust that will be the deciding factor,’ says
Gavrila. ‘The introduction of self-driving vehicles will therefore be a
phased process.’
Professor
Gavrila expects that by 2020 self-driving vehicles on highways will
enable drivers to keep their eyes off the road for prolonged periods of
time, at speeds of 100 km/h
and higher. This would allow drivers to use their smartphones, for
example. There are currently already hundreds of Waymo (Google) and Uber
cars operating as ‘robot taxis’ in the streets of Pittsburgh and
Phoenix in the US – although they still require back-up drivers. The
first large-scale tests without back-up drivers will be take place
within the next 2-3 years in simplified conditions. Gavrila expects that
it will take at least till 2030 before robot taxis – safer than
manually driven cars yet with a ‘socially acceptable’ driving style -
will be widespread in cities like Amsterdam. ‘By that time, self-driving
vehicles will have revolutionized our mobility.’
Opening of the RADD lab with demo Professor Gavrila’s inaugural
address on Friday 23 June will be followed on Tuesday 27 June by another
event at TU Delft at which self-driving vehicles will be the centre of
attention. This concerns the opening of the Researchlab Automated
Driving Delft (RADD) in The Green Village on the TU Delft campus. The
lab will offer extensive scope for experiments involving automated
driving. The opening of the lab will be performed by Melanie Schultz van
Haegen, the Dutch Minister of Infrastructure and the Environment.
More information Inaugural address by Professor Dariu Gavrila: ‘The Intelligent Vehicles (R)evolution’
Date: Friday 23 June 2017, at 15.00
Venue: TU Delft Aula conference centre
The inaugural address can be followed live or viewed later, via Collegerama TU Delft
