Abstract: Performance of cleaner robot are frequently limited by the contradiction between the requirement from complex environment and its scarce computing resource. We proposed a low-overhead YOLO based object detection and a robust V-SLAM method. Aided with depth-vision module, we implemented a region division and room prediction algorithm that enables differentiated cleaning inside the house. This design has achieved success in terms of cleanliness, coverage rate by improving 36.7% and 4.19% respectively, and decreases the failure rate by 5%. Other pragmatic functions implemented on the robot include home modeling and patrol, lost property recovery, alarm for gas leakage, etc.