We are looking forward to the International Conference on Machine Learning, 2022 #ICML where Nicholas Polosky, the lead author will be presenting our work on Wednesday (20th). Please join him if you are planning to participate this year.
As many of you are aware, an area of expertise in our MR Lab is applying machine learning for vision-based UAV autonomy. We have been developing some of these capabilities for our Naval Air Systems Command (NAVAIR) customer.
In such cases, when the risk associated with the agents learning online is high, we have to resort to offline learning with an emphasis on applying safety-related constraints. In such deployment scenarios, a robust guarantee on the cost of a policy is paramount in real-world scenarios where breaking a cost budget may carry extreme consequences. To address some of the associated challenges, in this paper, we introduce Constrained Offline Policy Optimization (COPO).
N. Polosky, BC. da Silva, M. Fiterau, J. Jagannath, “Constrained Offline Policy Optimization”, in Proc. of the 39th International Conference on Machine Learning (ICML), Baltimore, MD, July 2022
Reach out of you are interested in learning more about our #autonomy solutions for #unmanned systems!
For more information, visit: Getting Started (icml.cc)