WS Uncertainty in Open-World Robotics

leveraging uncertainty to enable safer, more adaptable, and intelligent robotic systems in dynamic, real-world environments

space robotics

autonomous driving

underwater robotics

household robotics

agriculture & forestry

robot learning

Can robots learn to say 'I don't know'?

Is confidence always a good thing?

How can we quantify the unknown?

What's open-world and closed-world segmentation?

Does more data always reduce the uncertainty?

How can uncertainty guide exploration?

What’s the cost of ignoring uncertainty?

Uncertainty is an important concept for robotics because it affects how robots perceive, navigate, and interact with the world. When considering an open-world setting, robots must not only make decisions but should also know how confident they are in those decisions. In perception, handling uncertainty is critical, especially in an open-world setting where robots encounter unfamiliar objects or scenes. Without an awareness of uncertainty, a robot might misclassify unknown elements, leading to potentially unsafe actions. In navigation, estimating uncertainty helps assess traversability, since robots need to be able to judge which paths are safer, even when dealing with incomplete data or changing terrain. Ignoring uncertainty can lead to overly cautious behavior or dangerous decisions. For control, uncertainty-aware systems can better handle typical problems such as sensor noise, unpredictable disturbances, and modeling errors, leading to more reliable and stable performance.

As robotics moves into high-stakes applications like autonomous driving and field robotics, robots need to manage uncertainty to aim for safe and reliable real-world deployment. This workshop aims to bring together researchers tackling uncertainty across various domains in open-world robotics. Our focus is on addressing uncertainty from a practical perspective, aiming to develop methods that can be applied to real-world robotic systems operating in dynamic and unstructured environments. Rather than treating uncertainty as a purely academic problem or a theoretical challenge, we emphasize its importance in enabling robots to make safer, more informed decisions in real-time scenarios. Through this workshop, we hope to foster collaboration and innovation that push the boundaries of what is possible in open-world robotics.