BOLT The Behavior Optimization and Learning Theory Laboratory

Research

2026

MINav s two-hour workflow: autonomous exploration, offline policy training on a laptop, and robot deployment.
120 Minutes and a Laptop: Minimalist Image-goal Navigation via Unsupervised Exploration and Offline RL
Xiaoming Liu, Borong Zhang, Qingbiao Li, Steven Morad
arXiv preprint  ·  27 Mar 2026  ·  https://arxiv.org/abs/2603.26441
Explore to Learn: four disentangled muscle synergy patterns for an 80-dimensional musculoskeletal control task.
AAAI 2026 · Oral · Top 5% Explore to Learn: Latent Exploration Through Disentangled Synergy Patterns for Reinforcement Learning in Overactuated Control
Yiming Wang, Kaiyan Zhao, Xu Li, Yan Li, Jiayu Chen, Steven Morad, Leong Hou U
Proceedings of the AAAI Conference on Artificial Intelligence  ·  14 Mar 2026  ·  doi:10.1609/aaai.v40i31.39876
Mixture density learning: NLL mode collapse compared with the separated modes learned by nGEM.
Learning Mixture Density via Natural Gradient Expectation Maximization
Yutao Chen, Jasmine Bayrooti, Steven Morad
arXiv preprint  ·  11 Feb 2026  ·  https://arxiv.org/abs/2602.10602
POPGym Arcade: paired fully and partially observable game environments and the memory analysis framework.
ICML 2026 · Spotlight · Top 2.2% Investigating Memory in Model-Free RL with POPGym Arcade
Zekang Wang*, Zhe He*, Borong Zhang, Edan Toledo, Steven Morad
International Conference on Machine Learning (ICML 2026)  ·  https://arxiv.org/abs/2503.01450

2025

Five real robots following language-conditioned navigation commands, with color-coded goals and trajectories after coordinating a three-way yield.
Language-Conditioned Offline RL for Multi-Robot Navigation
Steven Morad, Ajay Shankar, Jan Blumenkamp, Amanda Prorok
2025 IEEE International Conference on Robotics and Automation (ICRA)  ·  19 May 2025  ·  doi:10.1109/ICRA55743.2025.11127288
CoViS-Net controls three robots in formation using visual relative-pose estimates, with cyan, magenta, and yellow fields of view.
CoViS-Net: A Cooperative Visual Spatial Foundation Model for Multi-Robot Applications
Jan Blumenkamp, Steven Morad, Jennifer Gielis, Amanda Prorok
Conference on Robot Learning (CoRL 2024), PMLR 270  ·  12 Jan 2025  ·  https://proceedings.mlr.press/v270/blumenkamp25a.html

2024

Memoroids: trajectory-based batching and episode indices, Figure 2.
Recurrent Reinforcement Learning with Memoroids
Steven Morad, Chris Lu, Ryan Kortvelesy, Stephan Liwicki, Jakob Nicolaus Foerster, Amanda Prorok
Advances in Neural Information Processing Systems  ·  26 Sep 2024  ·  https://openreview.net/forum?id=nA4Q983a1v

2023

Fast and Forgetful Memory: exponential forgetting and oscillatory temporal context, Figure 1.
Reinforcement Learning with Fast and Forgetful Memory
Steven Morad, Ryan Kortvelesy, Stephan Liwicki, Amanda Prorok
Advances in Neural Information Processing Systems  ·  22 Sep 2023  ·  https://openreview.net/forum?id=KTfAtro6vP
Generalised f-Mean Aggregation: six aggregation-function surfaces from Figure 1.
Generalised f-Mean Aggregation for Graph Neural Networks
Ryan Kortvelesy, Steven Morad, Amanda Prorok
Advances in Neural Information Processing Systems  ·  22 Sep 2023  ·  https://openreview.net/forum?id=JMrIeKjTAe
PISA set autoencoder architecture: set encoder and set decoder from Figure 1.
Permutation-Invariant Set Autoencoders with Fixed-Size Embeddings for Multi-Agent Learning
Ryan Kortvelesy, Steven Morad, Amanda Prorok
International Conference on Autonomous Agents and Multiagent Systems  ·  30 May 2023  ·  https://arxiv.org/abs/2302.12826
POPGym: two maze navigation environments used to benchmark partially observable reinforcement learning.
POPGym: Benchmarking Partially Observable Reinforcement Learning
Steven Morad, Ryan Kortvelesy, Matteo Bettini, Stephan Liwicki, Amanda Prorok
The Eleventh International Conference on Learning Representations  ·  01 Feb 2023  ·  https://openreview.net/forum?id=chDrutUTs0K

2022

Five DJI RoboMaster robots coordinating through a narrow passage using decentralized GNN policies.
A Framework for Real-World Multi-Robot Systems Running Decentralized GNN-Based Policies
Jan Blumenkamp, Steven Morad, Jennifer Gielis, Qingbiao Li, Amanda Prorok
2022 International Conference on Robotics and Automation (ICRA)  ·  23 May 2022  ·  doi:10.1109/ICRA46639.2022.9811744

2021

NavACL ground robot navigating through an unfamiliar cluttered house toward a vase target.
Embodied Visual Navigation With Automatic Curriculum Learning in Real Environments
Steven D. Morad, Roberto Mecca, Rudra P. K. Poudel, Stephan Liwicki, Roberto Cipolla
IEEE Robotics and Automation Letters  ·  01 Apr 2021  ·  doi:10.1109/LRA.2020.3048662

2020

Visible-light and near-infrared camera rig beside a firn wall at Mount St. Helens.
Improving Visual Feature Extraction in Glacial Environments
Steven D. Morad, Jeremy Nash, Shoya Higa, Russell Smith, Aaron Parness, Kobus Barnard
IEEE Robotics and Automation Letters  ·  01 Apr 2020  ·  doi:10.1109/LRA.2019.2959490

2019

SPEER spherical microbot rendering with its shell, thruster, cameras, batteries, and onboard electronics labeled.
A Spring Propelled Extreme Environment Robot for Off-World Cave Exploration
Steven D. Morad, Thomas Dailey, Leonard Dean Vance, Jekan Thangavelautham
2019 IEEE Aerospace Conference  ·  01 Mar 2019  ·  doi:10.1109/AERO.2019.8741609