The recent success of foundation models has initiated a paradigm shift in robotics. The prospect of a general-purpose robot, capable of understanding natural language and adapting to new tasks with minimal training, seems closer than ever. This progress is largely fueled by scaling up model size and training data. While this approach has yielded impressive results, it has also created a looming sustainability and efficiency crisis. The carbon footprint of training a single large model is substantial, and the energy required to deploy these models on physical robots is often prohibitive. This not only raises environmental concerns but also limits the application of these advanced AI systems, especially in mobile robotics, where power is a finite resource.

This workshop is both timely and critical. It addresses a growing concern within the community that has not yet been the focus of a major workshop at ICRA. By concentrating on the intersection of sustainability, efficiency, and large-scale embodied AI, we aim to move the conversation beyond just capability and performance. We want to ask how we can achieve the next level of robotic intelligence in a way that is environmentally responsible, economically viable, and accessible to the broader research community.

Our workshop will be the first to assemble a dedicated forum at ICRA to discuss these challenges. The event will expand the diversity of content at the conference by highlighting an emerging and vital area of research. We believe the discussions and collaborations fostered here will have a lasting impact, encouraging the development of new benchmarks, methodologies, and hardware that prioritize efficiency and sustainability alongside performance.