Evangelos A. Theodorou

About

Evangelos A. Theodorou is a prominent researcher at the intersection of reinforcement learning, optimal control, and robotics, whose work has fundamentally advanced how machines learn complex motor behaviors. His most influential contributions center on path integral methods for reinforcement learning — a mathematically elegant framework that bridges stochastic optimal control and modern machine learning. His seminal 2010 paper, "A Generalized Path Integral Control Approach to Reinforcement Learning" (449 citations), introduced scalable algorithms that dramatically reduced computational overhead compared to classical approaches, while his companion work on high-dimensional motor skill learning (257 citations) demonstrated these methods' practical power in continuous, complex state-action spaces. Theodorou has made equally significant contributions to variable impedance control in robotics, drawing inspiration from biological motor control to enable robots to adaptively modulate stiffness and compliance under uncertainty — work that has collectively accumulated hundreds of citations. His research on learning from demonstration, movement segmentation, and robust grasping under uncertainty further illustrates a cohesive vision: robots that learn versatile, human-like manipulation skills autonomously. His robust Kalman filter work (143 citations) reflects his breadth, extending into sensor fusion and reliable perception. Collectively, Theodorou's research has profoundly shaped modern robot learning and motion planning.

Research Focus

Key Achievements

Top Papers

1. 1
   A Generalized Path Integral Control Approach to Reinforcement Learning

   449 citations · 2010
2. 2
   Learning variable impedance control

   329 citations · 2011
3. 3
   Reinforcement learning of motor skills in high dimensions: A path integral approach

   257 citations · 2010
4. 4
   Skill learning and task outcome prediction for manipulation

   207 citations · 2011
5. 5
   A Kalman filter for robust outlier detection

   143 citations · 2007
6. 6
   Learning to grasp under uncertainty

   88 citations · 2011
7. 7
   From dynamic movement primitives to associative skill memories

   71 citations · 2012
8. 8
   Movement segmentation using a primitive library

   71 citations · 2011
9. 9
   Variable Impedance Control - A Reinforcement Learning Approach

   68 citations · 2010
10. 10
    Model-Free Reinforcement Learning of Impedance Control in Stochastic Environments

    64 citations · 2012

Key Collaborators

Contact & Links