Mahdis

Mahdis Rabbani

She/her/hers
  • Mahdis’s research develops real-time game-theoretic planning and control methods for interaction-aware trajectory planning and decision-making in non-cooperative multi-agent systems. A central theme of her work is utilizing dynamic games to model strategic interaction: each agent optimizes its own objective while anticipating that other agents will do the same, and the resulting operating point is a Nash equilibrium. She designs computationally efficient methods to approximate these Nash solutions and embed them in a receding-horizon/MPC framework, enabling closed-loop trajectory planning that explicitly accounts for mutual influence among agents in uncertain, interactive environments. In parallel, she explores data-driven modeling of interaction and system dynamics (including Koopman- and learning-based approaches) to improve prediction and scalability beyond hand-crafted models. Her work is motivated by and demonstrated in autonomous driving and racing, with methods that generalize to broader interactive robotics settings.
Education and Degree(s)
  • Ph.D. Candidate in Mechanical Engineering at University of California, Davis
  • M.Sc. in Mechanical Engineering at University of California, Davis
  • B.Sc. in Mechanical Engineering at University of Tehran
Publications
  • Predictive Compensation in Finite-Horizon LQ Games under Gauss-Markov Deviations
  • Optimal Modified Feedback Strategies in LQ Games under Control Imperfections
  • Multi-Step Deep Koopman Network (MDK-Net) for Vehicle Control in Frenet Frame
  • Powertrain Hybridization for Autonomous Vehicles: Fuel Efficiency Perspective in Mixed Autonomy Traffic S Nazari, N Gowans, M Abtahi, M Rabbani - IEEE Transactions on Transportation Electrification, 2024
  • An Automatic Tuning MPC with Application to Ecological Cruise Control M Abtahi, M Rabbani, S Nazari - IFAC-PapersOnLine, 2023