Zahra Marvi

    • Postdoctoral associate, Mechanical engineering, University of Minnesota
    • Ph.D. Electrical Engineering Michigan State University
    • M.Sc. Mechatronics Engineering, K. N. Toosi University of Technology
    • B.Sc. Electrical Engineering, K. N. Toosi University of Technology
    • Research interest: control, Reinforcement learning, multi-agent systems, robotics.


Current and Past Work

With tremendous acceleration in electrified mobility (e-mobility), high-level coordination and control and getting the best out of the energy system are crucial. However, multi-domain safety-critical control, and system complexity make it an inherently challenging problem. As a postdoctoral associate, I am currently investigating the design of intelligent control and safe coordination of energy systems inspired by multi-agent techniques.

My research focus during my Ph.D. was to design data-driven, safe, and optimal controllers in the presence of model and environmental uncertainty to enable autonomy in dynamic systems. I developed safe and data-efficient off-policy reinforcement learning algorithms for safety-critical systems using control barrier functions (CBF). To address the safety challenge in exploratory data collection, I proposed employing an efficient learning framework with a prescribed performance in conjunction with a robust formulation to ensure the system’s safety and stability even during data collection and learning. To tackle the safety problem in the face of environmental uncertainty, I proposed a safety-aware model-learning and a novel CBF capable of safety guarantee with simplified modeling of the environment. In addition, to benefit from the cooperation of agents in an uncertain environment, I developed a cooperative framework for collision avoidance in multi-agent systems operating under heterogeneous measurement uncertainty by deploying information-gap theory.



[1] Marvi, Z, Kiumarsi, B. “Safe reinforcement learning: A control barrier function optimization approach”. Int J Robust Nonlinear Control. 2021; 31: 1923– 1940.

[2] Marvi, Z, Kiumarsi, “Barrier-certified learning-enabled safe control design for systems operating in uncertain environments,” IEEE/CAA J. Autom. Sinica, vol. 9, no. 3, pp. 1–13, Mar. 2022.

[3] Marvi, Z, Kiumarsi, B. “Robust Satisficing Cooperative Control Barrier Functions for Multi-Robots Systems using Information-Gap Theory”. Int J Robust Nonlinear Control. 2022.

[4]Marvi and B. Kiumarsi, “Reinforcement Learning with Safety and Stability Guarantees During Exploration for Linear Systems,” in IEEE Open Journal of Control Systems, vol. 1, pp. 322-334, 2022.


[5] Marvi and B. Kiumarsi, “Safety Planning Using Control Barrier Function: A Model Predictive Control Scheme,” 2019 IEEE 2nd Connected and Automated Vehicles Symposium (CAVS), Honolulu, HI, USA, 2019, pp. 1-5.

[6] Marvi and B. Kiumarsi, “Safe Off-policy Reinforcement Learning Using Barrier Functions,” 2020 American Control Conference (ACC), Denver, CO, USA, 2020, pp. 2176-2181.

[7] Marvi and B. Kiumarsi, “Barrier-certified Learning-based Control of Systems with Uncertain Safe Set,” 2021 American Control Conference (ACC), New Orleans, LA, USA, 2021.

[8] Marvi and B. Kiumarsi, “Barrier-Certified Model-Learning and Control of Uncertain Linear Systems using Experience Replay Method”, 2021 Conference on Decision and Control (CDC), Austin, TX, USA, 2021.

[9] Marvi, B. Kiumarsi, H. Modares, “High-confidence Barrier-certified Control Design using Goal-oriented Scenario Optimization and Experience Replay Model Learning”, 2022 (Under Review).