Jingqi Li
Hello! I'm Jingqi Li, a Peter O'Donnell Jr. Postdoctoral Fellow at the Oden Institute for Computational Engineering and Sciences, UT Austin. I work with Prof. David Fridovich-Keil on game-theoretic control and learning. I received my Ph.D. in Electrical Engineering and Computer Sciences (EECS) from UC Berkeley in 2025, where I was advised by Prof. Claire Tomlin and Prof. Somayeh Sojoudi.
Research interests:
As we deploy more autonomous systems, they will inevitably interact with other independent decision-makers. This raises a fundamental question: How should agents operating in the physical world reason, learn, and make strategic decisions under limited information?
My research addresses this question through the lenses of game theory, control, optimization, and reinforcement learning. Broadly, I study multi-agent decision-making under uncertainty, focusing on differentiable dynamic games (whose equilibria are differentiable with respect to game parameters), information structure, learning, and safety, as well as the computational methods that enable these systems to scale in practice. My work is motivated by applications in autonomous driving, drone racing, manipulation, advanced air mobility, and infrastructure-scale autonomy.
Current directions:
Differentiable Dynamic Games for Strategic Decision-Making: Optimization and learning-based approaches for strategic decision-making, leveraging differentiable structure and principled approximations for efficient computation and decision analysis;
Inference and Learning under Information Asymmetry: In practice, agents' uneven access to knowledge about their environment—information asymmetry—can disrupt coordination and lead to failures. Beyond seeking to align such asymmetry, I study how structured information asymmetry can instead shape strategic interaction and be harnessed to steer coordination outcomes;
Decentralized, Safe, Strategic Autonomy: Developing theoretical and algorithmic foundations for decentralized strategic autonomy, enabling agents to make safety-assured decisions, learn, and coordinate under limited information, with hardware robotics validation.
Selected works:
[SIOPT '24] We developed a differentiable approximation of the optimality conditions for constrained nonlinear feedback Stackelberg games with continuous action spaces. By approximating feedback game equilibrium conditions as differentiable equations subject to feedback best-response constraints, we propose a provably convergent inexact Newton method to solve constrained feedback nonlinear games.
[AAMAS '23] By differentiating through and inverting the map from objectives to feedback Nash equilibria, we proposed a tractable likelihood-gradient approximator for inverse feedback nonlinear games, facilitating intent inference under information asymmetry.
[RA-L '25] Leveraging Lipschitz continuity, we developed a reachability learning framework offering deterministic safety assurance under bounded uncertainties, extending beyond the probabilistic safety guarantees of prior works. Our method was verified in drone racing hardware experiments. [video]
I'm on the faculty job market this cycle (2027 fall) [CV]