The development of intelligent robots capable of accurately recognizing objects through vision and touch has made significant progress in recent years. By using tactile information obtained through sensors and machine learning algorithms, robots are now able to identify objects that they have previously handled. However, challenges still exist when it comes to distinguishing between objects that are similar in size and shape, or objects that are unknown to the robot. Factors such as background noise and variations in the shapes and sizes of the same type of object can also restrict robot perception.
Researchers from Tsinghua University have been working on overcoming these challenges by focusing on robotic recognition of common yet complex items. The team aimed to mimic human thermal sensing, which allows us to perceive hot and cold sensations, as well as distinguish between different materials based on their cooling properties. By designing a robotic tactile sensing method that incorporates thermal sensations, the researchers hoped to improve object detection accuracy and robustness.
The team developed a layered sensor capable of detecting material types at the surface, pressure sensitivity at the bottom, and thermal changes in a porous middle layer. They paired this sensor with an efficient cascade classification algorithm that can rule out object types in order of difficulty, starting with simpler categories like empty cartons before moving on to more complex objects like orange peels or scraps of cloth. This approach enabled the robot to perceive multiple attributes of the grasped object simultaneously, including thermal conductivity, thermal diffusivity, surface roughness, contact pressure, and temperature.
To test the capabilities of their method, the researchers created an intelligent robot tactile system designed to sort garbage. The robot was able to accurately identify and categorize a variety of common trash items, such as empty cartons, plastic bags, plastic bottles, napkins, sponges, orange peels, and expired drugs. The system achieved an impressive classification accuracy of 98.85% in recognizing diverse garbage objects that it had not encountered before. This successful garbage sorting behavior has the potential to reduce human labor in real-life scenarios and could be applied in various smart life technologies.
Future research in this area will focus on enhancing robotic embodied intelligence and autonomous implementation. The researchers also hope to explore the potential of combining their sensor technology with brain-computer interface technology, allowing tactile information collected by the sensor to be converted into neural signals that can be interpreted by the human brain. This could potentially re-empower tactile perception capabilities for individuals with hand disabilities, providing new opportunities for improving quality of life through advanced technology.