Road sense

It’s an odd experience, driving without your feet touching the pedals. The car speeds up along straights and slows coming into bends; it knows where the speed bumps are; it even brakes for traffic lights, although they’re green. All I have to do is steer, and stop my feet heading instinctively for the brake and accelerator. Eventually, I cross my ankles underneath the driving seat.

The car doesn’t have the same driving style as me: too slow into some corners, too fast into others, and I’m quite glad there are no other cars on the road. This isn’t a coincidence: we’re at the Transport Research Laboratory’stest track near Bracknell on a drizzly, overcast day. Occasionally, the intelligent driving system, called Sentience, fails; the engineers blame it on the cloud cover and the trees lining the track, which block the satellite signals the car is using to gauge its position. But despite this, there can be no doubt: it’s not fast and it’s not spectacular, but the car is quite definitely driving itself.

Sentience is the fruit of a large collaboration. Over the past 15 months, TRL, Jaguar Land Rover, Ricardo, Ordnance Survey and Orange have been working together to develop the system, which is designed to drive a hybrid vehicle in the most efficient way possible. The system brings a battery of technologies together to attack this problem: mobile communications and mapping combine to extend the vehicle’s ‘electronic horizon’; the cruise control system is co-opted to ensure the vehicle stays within speed limits; and the hybrid powertrain makes the best use of opportunities for battery charging. Even the air conditioning is tweaked to make the best use of energy.

With a system with so many elements, cooperation between the partners was vital. ‘Sentience brought together three industries that would not normally work together: automotive, telecommunications and mapping,’ said Tom Robinson of Ricardo, project director. It requires levels of intelligence throughout the system — data received via the GPS signals has to be processed in such a way that the car’s control systems can understand it and utilise it.

The centre of the system is Ricardo’s rCube controller, which acts as the supervisory control for the system’s inputs and outputs; analogous to the server on a computer network. This receives data via a Sentience application on a mobile phone — the developers used a Nokia N95 — which, in turn, gathers information from an external GPS receiver. Both of these were developed by Orange, with the phone also acting as the system’s human-machine interface (HMI).

Ordnance Survey, meanwhile, provided enhanced mapping for the navigation system. This comes into play when a route is entered into the navigation system, and at the moment is limited only to the TRL test track and a few routes on the roads in and around Cambridge.

The data is far finer-grained than the information normally conveyed by mapping systems, and is gathered by driving along the routes using GPS, video capture and logging techniques on a laptop. It includes such features as the elevation and compass bearing of the road ahead, the sharpness of the bends along the route, the gradients and cambers, the speed limits in force and the actual estimated speed of traffic, and street features that affect driving speed: junctions, roundabouts, schools, traffic calming measures and so on. ‘This sort of enhanced data is becoming increasingly available on a large scale,’ said Robinson; for the system to be truly practical, it would have to cover the entire area over which the vehicle would be used. In use, the driver would enter the postcode of his or her destination, and the extended mapping information would be downloaded into the navigation system, giving the control system full knowledge of all the geographical and street features between the start and end points.

All this information is fed into the rCube. Installed on the Sentience test vehicle — a Ford Hybrid Escape SUV, provided by Jaguar Land Rover — the rCube effectively hacks into the car’s electronic control system. This creates a number of new operating modes, Robinson said. Cruise control becomes Enhanced Acceleration and Deceleration, an overlay to the existing cruise control, which controls vehicle speed as well as the acceleration and deceleration profiles. This speeds the car up and slows it down between a number of defined set points on the route, including ‘fixed’ features such as speed limits, bends, roundabouts, give-way points and ‘probabilistic speed limits’ — areas where there is a strong chance that the car will have to slow down — pedestrian crossings, schools at certain times of day, junctions and traffic lights. ‘With the knowledge of future traffic and geography features, speed can be varied automatically at a closer to optimal rate than you would expect through a normal driving style,’ Robinson said. ‘That allows potentially significant savings in fuel.’

The system also takes control of the air conditioning. The idea is that the temperature in the cabin remains at a comfortable level while running the air conditioning compressor for a minimal time. In particular, it avoids running the compressor while the car is stationary; it does this by varying the temperature setpoints so that, when the system sees that the car is likely to stop, it drops the temperature below the setpoint. While the car is stopped, the temperature slowly rises, but doesn’t rise above the setpoint before the engine sets off again. ‘We do have a problem developing the system in the UK, though,’ Robinson said. ‘During the research-and-development phase, the weather never got warm enough for long enough to test it properly.’

Despite the spooky mind-of-its-own functioning of the EAD system, the real raison d’être of the Sentience is the Optimised Engine Loading (OEL) function. This allows the rCube to hack into the powertrain controller, which decides when the hybrid will use its petrol engine, when it will use its electric motors, and when it will use a combination of both. With recharging options available via the petrol engine and regenerative braking, this reduces the amount of fuel the car uses; charging on hills and at junctions as the vehicle brakes, adjusting the battery state of charge limits, and optimising the energy efficiency of the system. ‘Ford modified the software to provide us with the appropriate vehicle communications interfaces to allow the system to undertake supervisory control,’ said Robinson.

Sentience is clearly still within its demonstration and development phase, and not without glitches; a fully commercial version couldn’t be flummoxed by a cloudy day and some overhanging trees. Robinson estimates that it’s some four to five years away from commercialisation. However, Phil Pettitt, chief executive of the UK intelligent transport systems centre of excellence InnovITS, which co-funded the project, sees a great deal of potential for the system.

‘Because the acceleration and braking are smooth and gradual, it could act as a traffic-calming measure,’ he said. ‘It would help to keep traffic within the speed limits and damp down aggressive driving.’

Moreover, he said, the ‘automatic’ driving would help increase the mobility of elderly drivers, slowing the car down at possible trouble spots so that drivers don’t need to react so fast to hazards. ‘It would effectively increase the age range for drivers, so that people could keep their independence longer.’

Pettitt dismissed worries that the increased automation of driving de-skills drivers and could encourage lapses of concentration. ‘The system won’t come to a stop unless you actually put your foot on the brake, and it frees your mind so that you only have to concentrate on the conditions actually on the road, rather than the geographical features,’ he said. ‘After all, when synchromesh gears were invented and drivers didn’t have to double-declutch anymore, nobody was concerned that they didn’t have to concentrate so hard on the revs and the engine note.’