Conventional robot perception and navigation pipelines are built using traditional sensors such as RGB cameras, stereo depth sensors and LiDARs. These sensors scan the entire scene in a fixed and uniform way. In contrast, programmable light curtains are a recently-invented, resource-efficient sensor that measure the depth of any vertically-ruled surface (“curtain”) specified by the user. Compared to LiDARs, light curtains are relatively inexpensive, significantly faster (45-60 Hz) and capture depth at a much higher resolution (640 scan lines). However, they require user control.
The main contributions of this thesis are to (1) integrate programmable light curtains with an existing, state-of-the-art navigation and autonomy stack, (2) develop algorithms for enabling light curtains to detect and avoid obstacles for safe navigation, and (3) perform high resolution mapping and accurate robot localization using intelligent curtain placements. Our overall system consists of parallelized components that interact naturally and continuously while running at their own independent speeds. This work is a step towards full-stack autonomous robot navigation using fast, high-resolution, controllable sensing. We demonstrate our integration on a wheelchair robot.