Radar helmet could steer rescuers through smoke

日期:2017-07-11 08:00:16 作者:伯奖桦 阅读:

By Colin Barras (Image: Cinaz/Kenn) (Image: Cinaz/ Kenn) Rescue workers must sometimes navigate smoke-filled and unfamiliar buildings. But helmets packed with sensors and software designed to give robots a sense of direction could build instant maps to help with a mission. A technique called simultaneous location and mapping (SLAM) gives robots a version of the human ability to build up mental maps of an area as it is explored. The technique can be useful to, for example, let robotic cars navigate on unfamiliar streets. As a robot moves, distance sensors and cameras measure and record the distances to surrounding objects, making it possible to build a detailed and accurate map of the area. Burcu Cinaz at the University of Bremen in Germany and Holger Kenn at the European Microsoft Innovation Center, Achen, Germany, reasoned that the same technology could help humans navigate when visibility is reduced. “Under heavy smoke, neither [a rescue worker’s] own vision nor information captured by cameras is sufficient for localisation and mapping,” says Kenn. “However, there are technologies available that can penetrate smoke.” Radar and ultrasound are two examples. Cinaz and Kenn tested their idea by strapping an infrared laser scanner to a helmet, and wearing it while strolling down the corridor outside their office (see image, top right or this video (mp4 format, 3.8mb)). But the pair quickly encountered a problem. The SLAM software they had loaded onto a laptop connected to the helmet was designed for sensors gliding steadily along on wheels, and also needs a precise record of distances travelled to build an accurate map. But a walking person sways from side to side and their head bobs up and down, adding noise to the laser scanner signal. And there is no easy way to record the distance a person travels. “We tried to get away with a rather simple correction for head motion,” says Kenn. The researchers added an inertia-measuring device near to the laser scanner to record how much the person’s head swayed. This makes it possible to factor out confusing signals and have the laser scanner record a relatively steady stream of data. The revised version of the software also assumes a person is walking at a constant speed. The results are impressive – the software creates a map that accurately captures the shape of the area covered (see image, right). But so far the scale is always too small. Kenn thinks the solution could be to combine his approach with the work of Stéphane Beauregard at the University of Bremen. “He has developed more advanced motion and position estimation for pedestrians,” he says. They combine a head-mounted sensor with a second one mounted on the foot. Cinaz and Kenn presented their work at the Pervasive 2008 conference in Sydney,