Researchers led by Peeyush Thakur have built a navigation system for robotic nursing assistants that must move through cluttered rooms with low obstacles, according to a paper on arXiv. Lidar-based navigation maps rooms accurately but misses transparent or low obstacles, while camera-based systems suffer from noise and occlusion, the authors write.
The system uses the NavfnROS global path planner together with depth readings from an RGB-D camera and simple rules for avoiding obstacles. The researchers built it on a Husky A200 robot fitted with a Kinova Gen3 arm, which together form the nursing assistant.
In a simulated cluttered environment it achieved up to 40% higher success and 35% more minimum clearance from obstacles than setups with a single sensor, and the authors write that real-world experiments confirmed the approach.



