The project is a collaboration between innovITS ADVANCE and TRL and centres around a customisable mechanical dummy and post-analysis with high-speed video.
Pedestrian detection systems have been in development for a number of years and are starting to find their way into commercial vehicles — most notably as an optional extra on the Volvo C60. The technology is still maturing, however, and has difficulty operating at night or in poor weather conditions, for example.
‘To get it absolutely right, you should detect all people that you’re going to potentially collide with, but not end up triggering and emergency braking the vehicle just because there’s someone walking along at the side of the road in front of you or people crossing the road,’ said Phil Pettitt, chief executive at innovITS.
The automotive industry is largely lacking in dummy systems providing a realistic representation of human form and gait to the vehicle’s on-board sensor systems.
‘There’s a lot of complex, intelligent technologies getting into vehicles, and they need to work every time as expected, so you have to develop them carefully over several years and deal with every eventuality, which does mean extensive testing,’ Pettitt said.
‘I have seen presentations of pedestrian detection and the developers are the people walking in front of the vehicles. That’s a good statement of assurance but it’s not the best way — and you end up testing for one kind of person: essentially young engineers.’
The mechanical dummy developed by the team moves at preset speeds on a low-profile base unit that is invisible to vehicle systems. It can emulate the leg motion associated with a normal walking or running gait and can be configured for three body sizes: a 50th percentile adult male and adult female and a six-year-old child.
Each of these body options is fully detachable and designed to minimise vehicle damage should the pedestrian detection system fail to operate.
When combined with a unique ‘ground truth positioning system’ — based on high-definition, high-speed video — the pedestrian detection target system provides an accurate and repeatable testing capability.
‘You have an independent source of information about exactly what was happening in the test that you can then correlate with anything you’re logging in the vehicle as a result of the test,’ Pettitt said.
‘For instance, the two vehicles both thought they were 3m apart at the that point; looking at analysis of the video, we can see they’re actually 2.5m. That can help you understand why a system wasn’t working as you wanted it to.’