This dissertation presents several related algorithms that enable important capabilities forself-driving vehicles.
Using a rotating multi-beam laser rangefinder to sense the world, our vehicle scans millions of 3D points every second. Calibrating these sensors plays a crucial role in accurateperception, but manual calibration is unreasonably tedious, and generally inaccurate. Asan alternative, we present an unsupervised algorithm for automatically calibrating both theintrinsics and extrinsics of the laser unit from only seconds of driving in an arbitrary andunknown environment. We show that the results are not only vastly easier to obtain thantraditional calibration techniques, they are also more accurate.
A second key challenge in autonomous navigation is reliable localization in the faceof uncertainty. Using our calibrated sensors, we obtain high resolution infrared reflectivityreadings of the world. From these, we build large-scale self-consistent probabilistic lasermaps of urban scenes, and show that we can reliably localize a vehicle against these maps towithin centimeters, even in dynamic environments, by fusing noisy GPS and IMU readingswith the laser in realtime. We also present a localization algorithm that was used in theDARPA Urban Challenge, which operated without a prerecorded laser map, and allowedour vehicle to complete the entire six-hour course without a single localization failure.
Finally, we present a collection of algorithms for the mapping and detection of traffic lights in realtime. These methods use a combination of computer-vision techniquesand probabilistic approaches to incorporating uncertainty in order to allow our vehicle toreliably ascertain the state of traffic-light-controlled intersections.
猜您喜欢
推荐内容
开源项目推荐 更多
热门活动
热门器件
用户搜过
随便看看
热门下载
热门文章
热门标签
评论