How robots learn to recognize themselves
NEWS, Research, Robotics, Perception, Artificial Intelligence |
![](/fileadmin/_processed_/d/c/csm_Science_Robotics_31b32052fc.webp)
Please explain what you have found out. Why is it relevant for a robot to "learn elements of its morphology by itself"?
When designing a robot, we typically assume it already knows its physical structure, and any extra data from external devices is primarily used to adjust its movements. However, we are exploring a different concept: Can a robot autonomously acquire knowledge about its morphology? Picture the robot independently learning about its body, relying on minimal prior knowledge and utilizing only its proprioceptive sense – the sensation of position, movement, and effort. To explore this, we introduced a concept called "proprioceptive information graphs." These graphs depict the connections among various signals related to the robot's bodily structure. By analyzing these graphs, we aim to uncover crucial details about the robot, such as its body parts' connectivity and geometry. Our experiments with various robot types, including a robot arm, a six-legged robot, and a humanoid robot, demonstrate that we can effectively learn about their morphology using this approach, regardless of the number of parts or the specific shape of the robot.
Robots recognise their morphology autonomously
Where do you see practical scenarios for your research?
A robot's ability to autonomously acquire, monitor, and adapt knowledge of its morphology significantly enhances its performance across various scenarios due to heightened bodily awareness.
- In manipulation tasks, refined self-knowledge allows precise control of the robot's arm(s), resulting in more dexterous handling of objects. In the use of tools, the seamless integration of a tool as a temporary extension of the body morphology expands its applicability to diverse purposes.
- For robot locomotion, enhanced bodily awareness is crucial for navigating complex terrains and negotiating obstacles. Motion planning substantially benefits from continuously updated information about the robot's morphology.
- Safety and collaboration nuances arise from constant monitoring and adaptation of the robot's morphology, resulting in improved spatial awareness that helps avoid unintended contacts.
In summary, robots with the ability to autonomously understand and adapt to their morphology exhibit enhanced performance, precision in manipulation, expanded tool
utility, improved navigation, and increased safety awareness.
Weitere Informationen