Workshop on Robot Learning at Deployment (Learn@Deploy)

Robots deployed in the real world inevitably face novel variations, unseen tasks, and edge cases that no amount of offline pretraining can anticipate. Yet standard paradigms rely on static policies that cannot correct errors, self-critique, or optimize online.

This half-day workshop focuses on Robot Learning at Deployment Time (Learn@Deploy): the methods, safety guardrails, and evaluation strategies for continuous adaptation and test-time optimization after deployment, under real constraints on compute, memory, streaming data, and physical safety. We bring together industry and academic researchers across robot learning, embodied AI, planning, control, and computer vision to discuss the principles of safe, resource-bounded online adaptation, with an emphasis on interactive, small-group discussions.