Active mapping, passive localization

An active sensor sends out energy and measures what comes back. A lidar sends out laser beams, and a radar sends out radio signals. A passive sensor, such as a camera, only receives the light that is already in the scene.

Some applications, such as defence, require a platform with no emissions, because emissions make the platform easier to detect. Such a platform cannot use active sensors.

But even in defence, a platform does not have to be passive all the time. At some places and times, for example in peacetime, emissions are permitted. Our software can use all prior information when it localizes. In the example below, we build a SLAM map with a lidar, three cameras and an IMU. Then we localize in that map with only the passive sensors, the cameras, switched on. Because an active sensor saw the environment at an earlier time, the localization accuracy improves drastically, even though no active sensor is on at runtime.

Mapping with active sensors

The first video shows the mapping stage, at a time when emissions are permitted. A handheld device carries a 360° lidar, three cameras and an IMU. Our software fuses all of these sensors to estimate the path of the device (the odometry) and to build the SLAM map. Each lidar point takes its colour from the camera images, so the map shows the scene in its real colours.

Localization with passive sensors

The second video shows the localization stage. The lidar is off. We load the SLAM map and localize in it with the cameras only. The platform emits nothing, so the run is completely passive. The cameras still get the benefit of the lidar, because the map holds the geometry that the lidar measured.