Mirco Theile

Postdoctoral Researcher in Reinforcement Learning

Technical University of Munich (TUM)

I am a Postdoctoral Researcher at the Chair of Cyber-Physical Systems at the Technical University of Munich (TUM) advised by Marco Caccamo. Until recently I was also a visiting researcher with Alberto Sangiovanni-Vincentelli at UC Berkeley. My research is on Reinforcement Learning (RL) for Cyber-Physical Systems (CPS), with a main focus on UAV path planning for single and multi-agent systems. In other projects I am investigating physical safety and real-time constraints of deploying RL algorithms in the real world, and applying RL to DAG scheduling.

In July 2024, I graduated with a Dr. rer. nat. (~Ph.D.) in Computer Science from TUM, where I also received the M.Sc. degree in electrical engineering and information technology in 2018. During my M.Sc., I spent a semester at Royal Institute of Technology (KTH), Stockholm. For my Master Thesis I was a Visiting Scholar with the University of Illinois at Urbana–Champaign (UIUC) working on a long-endurance solar UAV. Before, I received my B.Sc. degree in electrical engineering and information technology from TU Dortmund in 2015.

Featured Publications

Equivariant Ensembles and Regularization for Reinforcement Learning in Map-based Path Planning

This paper describes a method to exploit environmental symmetries in reinforcement learning by constructing equivariant policies and invariant value functions without specialized neural network components. We show how equivariant ensembles and regularization benefit sample efficiency and performance in a map-based path planning case study.

Mirco Theile, Hongpeng Cao, Marco Caccamo, Alberto L. Sangiovanni-Vincentelli

Equivariant Ensembles and Regularization for Reinforcement Learning in Map-based Path Planning

Learning to Generate All Feasible Actions

We introduce action mapping, a novel approach that divides the learning process into two steps: first learn feasibility and subsequently, the objective by mapping actions into the sets of feasible actions.

Mirco Theile, Daniele Bernardini, Raphael Trumpp, Cristina Piazza, Marco Caccamo, Alberto L. Sangiovanni-Vincentelli

Learning to Generate All Feasible Actions

Edge Generation Scheduling for DAG Tasks using Deep Reinforcement Learning

This paper studies the problem of scheduling real-time DAG tasks by presenting a novel schedulability test based on the concept of trivial schedulability. Using this schedulability test, we propose a new DAG scheduling framework (edge generation scheduling – EGS) that attempts to minimize the DAG width by iteratively generating edges while guaranteeing the deadline constraint.

Binqi Sun, Mirco Theile, Ziyuan Qin, Daniele Bernardini, Debayan Roy, Andrea Bastoni, Marco Caccamo

Edge Generation Scheduling for DAG Tasks using Deep Reinforcement Learning

Learning to Recharge: UAV Coverage Path Planning through Deep Reinforcement Learning

To solve the power-constrained UAV coverage path planning problem with recharge, we propose a novel proximal policy optimization (PPO)-based deep reinforcement learning (DRL) approach with map-based observations, utilizing action masking and discount factor scheduling to optimize coverage trajectories over the entire mission horizon.

Mirco Theile, Harald Bayerlein, Marco Caccamo, Alberto L. Sangiovanni-Vincentelli

Learning to Recharge: UAV Coverage Path Planning through Deep Reinforcement Learning

Other Publications

Hongpeng Cao, Mirco Theile, Federico G. Wyrwal, Marco Caccamo (2022). Cloud-Edge Training Architecture for Sim-to-Real Deep Reinforcement Learning. In IROS.

PDF Cite Project Video DOI

Or D. Dantsker, Mirco Theile, Marco Caccamo (2022). A Cyber-Physical Prototyping and Testing Framework to Enable the Rapid Development of UAVs. In Aerospace.

PDF Cite Project DOI

Mirco Theile, Harald Bayerlein, Richard Nai, David Gesbert, Marco Caccamo (2021). UAV Path Planning using Global and Local Map Information with Deep Reinforcement Learning. In ICAR.

PDF Cite Code Project DOI

Harald Bayerlein, Mirco Theile, Marco Caccamo, David Gesbert (2021). Multi-UAV Path Planning for Wireless Data Harvesting With Deep Reinforcement Learning. In OJ-COMS.

PDF Cite Code Project DOI

Jonathan Ponniah, Mirco Theile, Or D. Dantsker, Marco Caccamo (2021). Autonomous hierarchical multi-level clustering for multi-uav systems. AIAA Scitech 2021 Forum.

Cite DOI

See all publications