Daniel Bashir avatar

Dexa/Daniel Bashir

Learn more

Data-Driven Robotics

Achieving success in robotics often hinges on the ability to gather extensive data, as highlighted by the need for 2700 hours of teleoperated robot movements. The innovative approach of hindsight relabeling with language allows for the effective training of language-conditioned policies, showcasing the potential for real-world applications. Despite the challenges, the progress made demonstrates that investing time and effort into data collection can lead to significant advancements in robotic capabilities.