Erik Berger

About

Erik Berger is a robotics researcher whose work sits at the intersection of physical human-robot interaction, machine learning, and motor skill acquisition. His research has made significant strides in enabling robots to detect, estimate, and respond to external perturbations during collaborative tasks — a critical challenge when robots must work safely and fluidly alongside human partners. Berger's most influential contribution applies Dynamic Mode Decomposition (DMD) to perturbation estimation in human-robot interaction, a 2015 paper that has garnered 124 citations and demonstrated that robots can identify and compensate for unexpected forces without relying on expensive dedicated force sensors. This line of work, reinforced by related studies using transfer entropy for low-cost sensor-based detection, reflects a pragmatic drive to make adaptive robot behavior accessible on real-world platforms. Beyond perturbation handling, Berger has pioneered kinesthetic bootstrapping — teaching humanoid robots motor skills through direct physical guidance — and developed imitation learning frameworks that capture the dynamic interplay between two interacting agents, extending naturally into responsive synthetic humanoids. His latent space policy search approach addresses the curse of dimensionality in reinforcement learning for high-degree-of-freedom robots. Collectively, his body of work represents a sustained and impactful effort to make robots genuinely collaborative partners in shared physical tasks.

Research Focus

Key Achievements

Top Papers

1. 1
   Estimation of perturbations in robotic behavior using dynamic mode decomposition

   124 citations · 2015
2. 2
   Kinesthetic Bootstrapping: Teaching Motor Skills to Humanoid Robots through Physical Interaction

   35 citations · 2009
3. 3
   Learning responsive robot behavior by imitation

   32 citations · 2013
4. 4
   Dynamic Mode Decomposition for perturbation estimation in human robot interaction

   27 citations · 2014
5. 5
   Inferring guidance information in cooperative human-robot tasks

   26 citations · 2013
6. 6
   Latent space policy search for robotics

   20 citations · 2014
7. 7
   Experience-based torque estimation for an industrial robot

   13 citations · 2016
8. 8
   Learning Two-Person Interaction Models for Responsive Synthetic Humanoids

   11 citations · 2024
9. 9
   Transfer entropy for feature extraction in physical human-robot interaction: Detecting perturbations from low-cost sensors

   10 citations · 2014
10. 10
    Towards a Simulator for Imitation Learning with Kinesthetic Bootstrapping

    10 citations · 2008

Key Collaborators

Contact & Links