Cleaning Robot Semantic Mapping for Object-Adaptive Task Control

The patent combines multiple types of sensors (camera, LIDAR, ultrasonic sensor, infrared sensor) into an integrated sensing system. This merging of different sensing modalities enables the cleaning robot to capture comprehensive environmental information, improving object identification accuracy while maintaining manageable system complexity through unified processing architecture.

The sensing system is designed with multi-functionality, where each sensor type serves multiple purposes: the camera captures visual information for object recognition, LIDAR provides depth mapping for navigation, ultrasonic sensors detect proximity obstacles, and infrared sensors identify thermal signatures. This universal approach allows a single sensor suite to perform diverse detection tasks, improving measurement precision without proportionally increasing device complexity.

The patent segments the obstacle avoidance function into multiple classification levels. Instead of treating all objects uniformly, the system first categorizes objects by type (e.g., furniture, debris, obstacles), then applies different avoidance or interaction strategies to each category. This segmentation enables nuanced responses: some objects are avoided, others are navigated around selectively, and certain objects may even be approached for cleaning tasks.

The obstacle avoidance behavior is made dynamic and adaptive rather than static and uniform. The cleaning robot continuously adjusts its avoidance strategy based on real-time object identification results from multiple sensors. When an object is identified as safe to approach (e.g., a designated cleaning area), the robot dynamically modifies its path planning to approach rather than avoid, demonstrating versatile task adaptation while maintaining operational simplicity through automated decision-making.

The system performs preliminary object identification and classification using multiple sensors before executing cleaning tasks. By pre-processing environmental data to identify object types, locations, and characteristics, the robot prepares structured information that guides subsequent task execution. This preliminary action reduces the processing burden during actual cleaning operations, as the robot already knows what to clean and how to approach it, thereby improving productivity without excessive energy consumption.

The cleaning robot autonomously processes and interprets sensor data to generate its own task plan without external intervention. The system self-services by automatically identifying objects, determining appropriate cleaning actions, and executing tasks based on its own sensory input and internal processing. This self-service capability improves productivity by eliminating delays associated with human monitoring and decision-making, while energy consumption is optimized through efficient onboard processing algorithms.