Uploaded February 2025 | Updated September 2026, 2 weeks ago
By watching their own motions with a camera, robots can teach themselves about the structure of their own bodies and how they move
Sequence 1
The data collection process for robots 1 and 2: this footage demonstrates the data collection phase, where a camera is focused on robot 1 to capture the RGB images. The recording mimics the concept of a robot observing itself in a mirror, allowing for the collection of essential data for the free form kinematic self-modelling.
Sequence 2
A 3D morphology prediction of robot arms: displaying the capabilities of our trained model, this segment showcases the prediction of 3D points occupied by robot arms based on motor action commands. The predictions are visually represented by a green shadow overlay on the robot arm, providing a clear and illustrative view of the model’s output.
Sequence
In this segment, we demonstrate the application of our robot arm in motion planning tasks. Referencing the ‘Motion planning using FFKSM’ section, it showcases how the robot utilizes its self-learned model to execute manipulation tasks, highlighting the practical implications and adaptability of our framework in real-world scenarios.
By watching their own motions with a camera, robots can teach themselves about the structure of their own bodies and how they move
Sequence 1
The data collection process for robots 1 and 2: this footage demonstrates the data collection phase, where a camera is focused on robot 1 to capture the RGB images. The recording mimics the concept of a robot observing itself in a mirror, allowing for the collection of essential data for the free form kinematic self-modelling.
Sequence 2
A 3D morphology prediction of robot arms: displaying the capabilities of our trained model, this segment showcases the prediction of 3D points occupied by robot arms based on motor action commands. The predictions are visually represented by a green shadow overlay on the robot arm, providing a clear and illustrative view of the model’s output.
Sequence
In this segment, we demonstrate the application of our robot arm in motion planning tasks. Referencing the ‘Motion planning using FFKSM’ section, it showcases how the robot utilizes its self-learned model to execute manipulation tasks, highlighting the practical implications and adaptability of our framework in real-world scenarios.
