Publications

Affordance-Based Hierarchical Reinforcement Learning for Quadruped Pedipulation

Tuba Girgin, Jose Castelblanco, Gabriel Rodriguez, Emre Girgin, Cagri Kilic

preprint — 2026

TL;DR A hierarchical reinforcement learning framework enables quadruped robots to autonomously choose where to stand and how to interact with objects, eliminating the need for hand-designed manipulation trajectories.

AcroRL: Learning Aggressive Quadrotor Inversion using Bidirectional Thrust

Gabriel Rodriguez*, Henri Sayag, Abhishek Rathod, John Stecklein, Siddharth Saha, Christopher Barngrover, Wennie Tabib

preprint — 2026

TL;DR A reinforcement learning framework that trains separate policies for compact, position-constrained quadrotor inversions using bidirectional thrust, reducing position RMSE by 32% and settling time by 57% over the best optimization-based baseline, with successful zero-shot hardware transfer.