Avatar

Ahalya Prabhakar

Lecturer and Associate Research Scientist

Yale University

Biography

I am a Lecturer and Associate Research Scientist at Yale University in the Department of Mechanical Engineering. My work focuses on developing learning and control algorithms for efficient human-robot interaction. Prior to Yale, I was a Postdoctoral Research Fellow in the Learning Algorithms and Systems (LASA) lab at the École Polytechnique Fédérale de Lausanne (EPFL) under Prof. Aude Billard, where I worked on developing algorithms for multimodal sensory learning for safe manipulation and adaptive safety controllers for human-robot collaboration. I obtained my Ph.D. in Robotics at Northwestern University in Prof. Todd Murphey’s lab. My Ph.D. work centered on developing information communication algorithms to enable intuitive human-robot collaboration and efficient robot learning using distribution-based motion representations and information-theoretic measures. I led Northwestern’s team for the DARPA OFFSET Urban Swarm Challenge, developing autonomous swarm algorithms for shared human-swarm collaboration and exploring their efficacy in aiding task performance under dynamic, time-sensitive constraints.

Interests

  • High-Dimensional Robot Learning
  • Information-Theoretic Motion Representations
  • Human-Robot Collaboration
  • Learning from Demonstrations

Education

  • PhD in Mechanical Engineering, 2020

    Northwestern University

  • MSc in Mechanical Engineering, 2016

    Northwestern University

  • BSc in Mechanical Engineering, 2013

    California Institute of Technology

Projects

High-Dimensional Learning for Robot Control

When learning representations for robot control, the quality of the learned model (whether of the world, task or system itself) depends …

Shared Ergodic Control for Flexible Human-Swarm Collaboration under Pressure

Developing autonomous ergodic swarm algorithms for human-swarm collaboration under dynamic, time-sensitive constraints

Learning Information-based Task Representations

People are extraordinarly good at crafting representations of tasks that takes into account the motion information necessary for …

Recent Publications

A Predictive Model for Tactile Force Estimation using Audio-Tactile Data

Robust in-hand manipulation of objects with movable content requires estimation and prediction of the contents’ motion with …

Mechanical intelligence for learning embodied sensor-object relationships

Intelligence involves processing sensory experiences into representations useful for prediction. Understanding sensory experiences and …

Scale-Invariant Specifications for Human-Swarm Systems

We present a method for controlling a swarm using its spectral decomposition—that is, by describing the set of trajectories of a swarm …

Credit Assignment Safety Learning from Human Demonstrations

A critical need in assistive robotics, such as assistive wheelchairs for navigation, is a need to learn task intent and safety …

Contact