Scalable Autonomy Through End-to-End Learning
We explore scalable approaches to automated driving by developing and evaluating end-to-end models that map sensor inputs directly to driving actions. This research pursues autonomy that is efficient, adaptive, and capable of operating safely in complex driving environments.
Data-Centric Development
Our team builds the infrastructure and pipelines needed to train advanced driving systems on large, diverse datasets. We focus on defining high-quality dataset formats, implementing data filtering and curation strategies, and integrating external visual-language sources like aerial imagery and traffic cameras.
Simulated and Real-World Domains
We rigorously test autonomous systems in a range of environments—from controlled closed courses to open-road settings. Virtual validation, such as sensor simulation and closed-loop testing, complements physical evaluations, which guide the development of increasingly safer and more capable systems.
