Lars Ankile

Papers

Total Citations

H-Index

About

Lars Ankile is a rising researcher at the forefront of robotic manipulation, specializing in data-efficient imitation learning and precision assembly. His work addresses a critical bottleneck in robotics: enabling robots to learn complex, long-horizon tasks from minimal human demonstrations. In his highly cited 2024 paper, "JUICER: Data-Efficient Imitation Learning for Robotic Assembly" (21 citations), Ankile introduced a novel pipeline that dramatically improves imitation learning performance without requiring large demonstration datasets, making precise manipulation more accessible. Building on this, his 2025 work "From Imitation to Refinement - Residual RL for Precise Assembly" (7 citations) identifies a fundamental limitation of Behavior Cloning—performance saturation with increasing data—attributing this to two critical factors. Ankile proposes a residual reinforcement learning framework that refines imitation-learned policies, achieving superior precision and reliability. His contributions are pivotal for advancing robotic assembly in manufacturing and automation, where both data efficiency and precision are paramount. As an emerging leader in robot learning, Ankile’s research bridges the gap between practical teachability and high-performance execution, with his work already shaping how researchers approach visuomotor policy learning for complex tasks.