Goal-Conditioned Learning MPC

May 1, 2023 · 1 min read
projects

Independent Research (May 2023 — Sep 2023)

  • Explored a goal-conditioned, data-driven MPC pipeline using feasible closed-loop trajectories to learn transition costs between feasible states via imitation learning.
  • Investigated a two-layer hierarchical architecture: short-horizon reference generation in a planning layer, tracked by MPC in an execution layer.
  • Demonstrated improved long-horizon efficiency in simulation compared to a single-layer MPC baseline.

Tech: Gurobi, CasADi, MPC, Python.

Model Predictive Control Imitation Learning Hierarchical Control
Shengfan Cao
Authors
Shengfan Cao (he/him)
PhD Researcher in Autonomous Driving & Robotics
UC Berkeley Ph.D. candidate developing safe, machine-learning-based prediction, planning, and control systems for autonomous vehicles. First-author IEEE IROS oral presenter with 4+ years across machine learning, planning, and controls, spanning PyTorch research, CARLA evaluation, embedded-robot deployment, and C++ production planner integration.