Ian Abraham

I am Post-doctoral fellow at The Robotics Institute at Carnegie Mellon University.

I obtained my Ph.D. at Northwestern University in the Center for Robotics and Biosystems. During my Ph.D. I interned at the NVIDIA Seattle Robotics Lab where I worked on model-based control for kitchen robotics. I was also a part of the DARPA FX-3 Urban Swarm Challenge where I implemented my decentralized ergodic control work on a swarm of field robots. I am also the recipient of the 2019 T-RO best paper award for my work on active learning for linear representations of nonlinear systems.

Email  /  Blog  /  twitter  /  CV  /  GitHub  /  Google Scholar

profile photo


My research is focused on the development of active runtime learning techniques for robotic systems. Ultimately, my goal is to enhance robotic systems to be self-sufficient and adaptive with minimal human intervention.


project image Hybrid Control for Learning Motor Skills
Ian Abraham, Alexander Broad, Allison Pinosky, Brenna Argall, Todd Murphey
Workshop on Algorithmic Foundations of Robotics , 2020

We develop a hybrid control approach for robot learning based on combining learned predictive models with experience-based state-action policy mappings to improve the learning capabilities of robotic systems. Hybrid control tools are used to create an algorithmic approach for combining learned predictive models with experience-based learning. Hybrid learning is presented as a method for efficiently learning motor skills by systematically combining and improving the performance of predictive models and experience- based policies.

project image An Ergodic Measure for Active Learning From Equilibrium
Ian Abraham, Ahalya Prabhakar, Todd Murphey
IEEE Transactions on Automation Science and Engineering, 2020

This paper develops KL-Ergodic Exploration from Equilibrium (KL-E 3 ), a method for robotic systems to integrate stability into actively generating informative measurements through ergodic exploration. Ergodic exploration enables robotic systems to indirectly sample from informative spatial distributions globally, avoiding local optima, and without the need to evaluate the derivatives of the distribution against the robot dynamics.

project image Model-based Generalization under Parameter Uncertainty using Path Integral Control
Ian Abraham, Ankur Handa, Nathan Ratliff, Kendall Lowrey, Todd Murphey, Dieter Fox
IEEE Robotics and Automation Letters, 2020
project page

This work addresses the problem of robot interaction in complex environments where online control and adaptation is necessary. By expanding the sample space in the free energy formulation of path integral control, we derive a natural extension to the path integral control that embeds uncertainty into action and provides robustness for model-based robot planning.

project image Active Area Coverage from Equilibrium
Ian Abraham, Ahalya Prabhakar, ToddD Murphey
Workshop on Algorithmic Foundations of Robotics, 2019

This paper develops a method for robots to integrate stability into actively seeking out informative measurements through coverage. We show that our method is able to maintain Lyapunov attractiveness while still actively seeking out data.

project image WHISKiT Physics: A three-dimensional mechanical model of the rat vibrissal array
NadinaO Zweifel, NickE Bush, Ian Abraham, ToddD Murphey, MitraJZ Hartmann
bioRxiv, 2019

Rodents tactually explore the environment using ~62 whiskers (vibrissae), regularly arranged in arrays on both sides of the face. The rat vibrissal system is one of the most commonly used models to study how the brain encodes and processes somatosensory information. To date, however, researchers have been unable to quantify the mechanosensory input at the base of each whisker, because the field lacks accurate models of three-dimensional whisker dynamics. To close this gap, we developed WHISKiT Physics, a simulation framework that incorporates realistic morphology of the full rat whisker array to predict time-varying mechanical signals for all whiskers.

project image Data-driven Koopman Operators for Model-based Shared Control of Human-Machine Systems
Alexander Broad, Ian Abraham, Brenna Argall, Todd Murphey
(conditionally Accepted) International Journal of Robotics Research, 2019

We present a data-driven shared control algorithm that can be used to improve a human operator’s control of complex dynamic machines and achieve tasks that would otherwise be challenging, or impossible, for the user on their own. Our method assumes no a priori knowledge of the system dynamics. Instead, both the dynamics and information about the user’s interaction are learned from observation through the use of a Koopman operator.

project image Active Learning of Dynamics for Data-Driven Control using Koopman Operators
Ian Abraham, Todd D. Murphey
IEEE Transactions on Robotics (King-Sun Fu Best Paper Award), 2019
project page code

This paper presents an active learning strategy for robotic systems that takes into account task information, enables fast learning, and allows control to be readily synthesized by taking advantage of the Koopman operator representation.

project image Decentralized ergodic control: distribution-driven sensing and exploration for multiagent systems
Ian Abraham, Todd D. Murphey
IEEE Robotics and Automation Letters, 2018

We present a decentralized ergodic control policy for time-varying area coverage problems for multiple agents with nonlinear dynamics. Ergodic control allows us to specify distributions as objectives for area coverage problems for nonlinear robotic systems as a closed-form controller.

project image Data-Driven Measurement Models for Active Localization in Sparse Environments
Ian Abraham, Anastasia Mavrommati, ToddD Murphey
Robotics: Science and Systems, 2018

We develop an algorithm to explore an environment to generate a measurement model for use in future localization tasks. Ergodic exploration with respect to the likelihood of a particular class of measurement (eg, a contact detection measurement in tactile sensing) enables construction of the measurement model. Exploration with respect to the information density based on the data-driven measurement model enables localization.

project image Structured neural network dynamics for model-based control
Alexander Broad, Ian Abraham, Todd Murphey, Brenna Argall
Learning and Inference in Robotics -- Robotics: Science and Systems, 2018

We present a structured neural network architecture that is inspired by linear time-varying dynamical systems. The network is designed to mimic the properties of linear dynamical systems which makes analysis and control simple. The architecture facilitates the integration of learned system models with gradient-based model predictive control algorithms, and removes the requirement of computing potentially costly derivatives online.

project image Real-Time Area Coverage and Target Localization using Receding-Horizon Ergodic Exploration
Anastasia Mavrommati, Emmanouil Tzorakoleftherakis, Ian Abraham and Todd D. Murphey
IEEE Transactions on Robotics, 2017

Although a number of solutions exist for the problems of coverage, search and target localization—commonly addressed separately—whether there exists a unified strategy that addresses these objectives in a coherent manner without being application-specific remains a largely open research question. In this paper, we develop a receding-horizon ergodic control approach, based on hybrid systems theory, that has the potential to fill this gap.

project image Ergodic exploration using binary sensing for nonparametric shape estimation
Ian Abraham, Ahalya Prabhakar, MitraJZ Hartmann, ToddD Murphey
IEEE Robotics and Automation Letters, 2017

Current methods to estimate object shape—using either vision or touch—generally depend on high-resolution sensing. Here, we exploit ergodic exploration to demonstrate successful shape estimation when using a low-resolution binary contact sensor.

project image Model-Based Control Using Koopman Operators
Ian Abraham, Gerardo De La Torre, Todd D. Murphey
Robotics: Science and Systems, 2017

This paper explores the application of Koopman operator theory to the control of robotic systems. The Koopman operator takes any observation of state at time t and evolves the functions of state subject to its dynamics forward in time linearly. Thus, a nonlinear dynamical system which has a discoverable Koopman operator can be represented by a linear Koopman operator in a lifted function space where the observations of state evolve linearly. We illustrate how the operator can be used to obtain a linearizable data-driven model of robotic systems for model-based control synthesis.

project image A nonlinear leg damping model for the prediction of running forces and stability
Ian Abraham, ZhuoHua Shen, Justin Seipel
Journal of Computational and Nonlinear Dynamics, 2015

Despite the neuromechanical complexity underlying animal locomotion, the steady-state center-of-mass motions and ground reaction forces of animal running can be predicted by simple spring-mass models such as the canonical spring-loaded inverted pendulum (SLIP) model. Here, we develop a leg damping model that is bilinear in leg length and velocity in order to yield improved damping force and ground reaction force prediction.

project image Model predictive control of buoyancy propelled autonomous underwater glider
Ian Abraham, Jingang Yi
ACC American Control Conference, 2015

Autonomous underwater gliders (AUG) is a cost-effective and efficient tool for oceanic exploration and discovery. In this paper, we present a model predictive control (MPC) design to compensate for the drift due to disturbances.

Design and source code from Jon Barron's website