Papers
7
Total Citations
95
H-Index
5
About
Oswin So is a robotics and control researcher whose work sits at the intersection of safe multi-agent systems, stochastic optimal control, and deep learning. His research addresses one of the most pressing challenges in autonomous robotics: enabling large numbers of agents to operate safely and efficiently in complex, uncertain environments without relying on centralized coordination. So's most recognized contributions center on scalable safe control for multi-robot systems. His GCBF+ framework introduced a neural graph control barrier function approach for distributed multi-agent safety, garnering 22 citations shortly after publication, while his 2024 survey on safe multi-robot control methods has already accumulated 32 citations, reflecting broad community interest. His earlier work explored variational inference for model predictive control using Tsallis divergence, offering a generalized probabilistic framework for stochastic optimal control (17 citations), and risk-sensitive planning through CVaR optimization. Beyond multi-agent systems, So has made meaningful contributions to stabilize-avoid optimal control problems, reformulating difficult multi-objective safety tasks via epigraph form deep reinforcement learning. His work on decentralized stochastic control using deep FBSDEs and ADMM further demonstrates his range across both theory and scalable implementation. Together, these contributions establish So as an emerging voice in principled, safety-aware autonomy research.
Research Focus
Key Achievements
Top Papers
- 1
- 2
- 3Variational Inference MPC using Tsallis Divergence17 citations · 2021
- 4
- 5
- 6
- 7