When detecting an obstacle, the robot vacuum can automatically adjust its angular speed and reduce its line speed to avoid the collision. The robot vacuum is designed by Kai-Tai Song, a professor in the Department of Electrical and Computer Engineering at National Chiao Tung University and his team. (Picture provided by Prof. Song)

Although robot vacuums have been widely used in many houses to keep the floor clean, the ones on the current market still have many problems. For example, robot vacuums may not clean the floor thoroughly, or they may hit on the obstacle before turning. The problems are solved by Kai-Tai Song, a professor in the Department of Electrical and Computer Engineering at National Chiao Tung University and his team. Collaborating with Quanta Computer Inc. in the project of "Path Planning for Robot Vacuums" for two years, Song''s team developed the technologies of "embedded image synchronization for positioning" and "map construction" to enable the positioning and navigating in robot vacuums. Their robot vacuums can automatically construct the environment map, not only for boosting the efficiency of cleaning but also for automatic obstacle avoidance, preventing any bump against the furniture.

According to Song, robots have been widely applied to the services in our houses and daily lives, and the robot vacuum has set up a great example. However, most of the current robot vacuums move randomly when cleaning up, and they turn only when hitting the obstacles, causing deficient cleaning. Thus, Song''s team added sensors and embedded the CPU on the robot, and develop a light-weight vSLAM algorithm to enable the automatic map construction and to optimize the cleaning path according to the environment map.

Song claims that the robot vacuums designed in NCTU can automatically plan the cleaning path which follows Zig-Zag, the most efficient cleaning path they found. Whenever the robot vacuums encounter unexpected obstacles, such as things, pets, or humans passing by, they can detect the obstacles with InfraRed, avoid the obstacles and then return to the planned path to resume the cleaning.

"The technology of automatic dodging will facilitate the application of self-driving cars and service robots," said Song. Future hospitals may replace human workers with robots for simple tasks, such as medicine delivery or document delivery. The automatic-dodging technology allows robots to carry out the missions themselves without hitting humans. He also stated that it is expected to see robots in the future campus walking or sweeping in the hallway or the classroom, and it is possible to design out-door cleaning robots with larger equipment.

Prof. Song''s robot vacuum has a camera, Realsense D435 RGB-D, providing both color and depth images with the resolution 640×480 and the frame rate 30fps. (Picture provided by Prof. Song)
 
Prof. Song''s robot vacuum plans the path in the shape of Zig-Zag after constructing a 2-dimensional grid map. (Picture provided by Prof. Song)