Lars Ankile

I'm a Research Assistant in Robotics at MIT CSAIL currently working on robotic manipulation and control. I previously worked with Professor David Parkes at Harvard studying multi-agent systems. My research focuses on developing algorithms that enable robots to learn complex manipulation skills. I'm particularly interested in combining imitation learning and reinforcement learning approaches to achieve reliable and precise robotic control. Some of my recent work has focused on diffusion-based policies for robotic manipulation, data-efficient imitation learning for assembly tasks, and scalable data collection systems for robot learning. Through my research, I aim to make robots more capable of performing real-world manipulation tasks by developing learning algorithms that can effectively leverage demonstrations, exploration, and large-scale data collection. I've published papers at major robotics and machine learning venues, with applications ranging from furniture assembly to precise object insertion tasks.

Check out videos and code for my projects at https://ankile.com

Publications

Robot Learning with Super-Linear Scaling

Robot Learning with Super-Linear Scaling

M. Torné, Arhan Jain, Jiayi Yuan, Vidaaranya Macha, L. Ankile, A. Simeonov, Pulkit Agrawal, Abhishek Gupta

DexHub and DART: Towards Internet Scale Robot Data Collection

DexHub and DART: Towards Internet Scale Robot Data Collection

Younghyo Park, Jagdeep Singh Bhatia, L. Ankile, Pulkit Agrawal

Diffusion Policy Policy Optimization

Diffusion Policy Policy Optimization

Allen Z. Ren, Justin Lidard, L. Ankile, A. Simeonov, Pulkit Agrawal, Anirudha Majumdar, Benjamin Burchfiel, Hongkai Dai, Max Simchowitz

arXiv.org 2024

From Imitation to Refinement -- Residual RL for Precise Assembly

From Imitation to Refinement -- Residual RL for Precise Assembly

L. Ankile, A. Simeonov, Idan Shenfeld, M. Torné, Pulkit Agrawal

JUICER: Data-Efficient Imitation Learning for Robotic Assembly

JUICER: Data-Efficient Imitation Learning for Robotic Assembly

L. Ankile, A. Simeonov, Idan Shenfeld, Pulkit Agrawal

arXiv.org 2024

Discovering User Types: Mapping User Traits by Task-Specific Behaviors in Reinforcement Learning

Discovering User Types: Mapping User Traits by Task-Specific Behaviors in Reinforcement Learning

L. Ankile, B. S. Ham, K. Mao, E. Shin, S. Swaroop, F. Doshi-Velez, W. Pan

arXiv.org 2023

Denoising Diffusion Probabilistic Models as a Defense against Adversarial Attacks

Denoising Diffusion Probabilistic Models as a Defense against Adversarial Attacks

L. Ankile, Anna Midgley, Sebastian Weisshaar

arXiv.org 2023

Deep Learning and Linear Programming for Automated Ensemble Forecasting and Interpretation

Deep Learning and Linear Programming for Automated Ensemble Forecasting and Interpretation

L. Ankile, Kjartan Krange

Deep Convolutional Neural Networks: A survey of the foundations, selected improvements, and some current applications

Deep Convolutional Neural Networks: A survey of the foundations, selected improvements, and some current applications

L. Ankile, Morgan Feet Heggland, Kjartan Krange

arXiv.org 2020

Approximate Strategy-Proofness in Large, Two-Sided Matching Markets

Approximate Strategy-Proofness in Large, Two-Sided Matching Markets

L. Ankile, Kjartan Krange, Yuto Yagi

arXiv.org 2019

The DONUT Approach to EnsembleCombination Forecasting

L. Ankile, Kjartan Krange

arXiv.org 2022

I See You! Robust Measurement of Adversarial Behavior

L. Ankile, David C. Parkes