A naukri.com initiative
Mit
1w
183
Image Credit: Mit
Teaching a robot its limits, to complete open-ended tasks safely
- Researchers at MIT’s Computer Science and Artificial Intelligence Laboratory used a combination of vision models and large language models (LLMs) to train robots to execute open-ended tasks safely and efficiently, ensuring machines deal with uncertainties and constraints. Their trial-and-error method, Planning for Robots via Code for Continuous Constraint Satisfaction (PRoC3S), creates a simulation on-the-fly of what’s around the robot and tests long-horizon plans to ensure they satisfy all constraints, until it arrives at one that can be executed. In simulations, PRoC3S successfully executed diverse tasks such as drawing, writing, and block placement, while also applying the approach to real-world tasks on robotic arms and table surfaces.
- The MIT researchers aim to use more scalable data-search techniques and applicable to mobile robots such as a quadruped for tasks that include walking and scanning surroundings. PRoC3S could provide assistance in public care centers and other areas requiring human-robot collaboration. The research team was supported, in part, by the National Science Foundation, the Air Force Office of Scientific Research, the Office of Naval Research, the Army Research Office, MIT Quest for Intelligence, and The AI Institute.
Read Full Article
11 Likes
For uninterrupted reading, download the app