Manipulating objects is something everybody has to learn. It's possible because our body recognizes patterns. Every child must first understand how to open a drawer. Afterwards, similarities between different situations are eventually recognized and what has been learned beforehand is applied. Machine Learning uses the same idea, making object manipulation through a robot fundamentally possible.
We believe that teaching a robot how to solve tasks and interact with objects should be an engaging challenge. Training will be done via an interface that makes use of Unity's Machine Learning Agents Toolkit tailored for robotics. Understanding and improving on the different algorithms that generate each motion sequence will be key for developing different interactions.
ROS2 is a middleware that has become the standard in robotics, specially for prototyping. Ubik's robotic arm comes with the necessary packages for easy access to all the variables used during the training process of your robot. Communicating through ROS2 also gives you the flexibility to add any sensors you want that might improve on the result.
Computer Vision makes 3D reconstruction and 6D pose estimation of objects possible in real time. We provide an integration of these models into the GUI through ROS2. By improving this models, the gap between simulation and real world environment can be continuously narrowed.
Once the robot has mastered a new skill, you can add it to the UI and relate it to the objects it was made for. Also you can collaborate with others and make improvements on it. This way the goal of robots eventually becoming every day helpers becomes a shared effort and an engaging challenge.
Free AI Website Creator