Deploying Reinforcement Learning in the Real World: A Case Study on Apptronik Apollo

Deep reinforcement learning (RL) has gained increasing popularity as an approach to achieving dynamic behaviors on legged robots. However, transferring RL behaviors from simulation to reality is a challenging process: imperfect sensors, simulation models, control architecture, and latency all present obstacles to successfully deploying such approaches in the real world. In this thesis, we present an end-to-end overview of our approach to bridging the sim-to-real gap, leveraging domain randomization and careful choices in control architecture in order to successfully deploy RL policies for teleoperation in simulation and on hardware.

Name thesis:degree_name

Master of Science

Level thesis:degree_level

masters

Discipline thesis:degree_discipline

Electrical Engineering

Department dc:contributor.department

Electrical Engineering

Grantor dc:publisher

Virginia Tech

Year dc:date.issued

2025

Author dc:creator

• Welch, Stephen Brian

Chair dc:contributor.committeechair

• Stilwell, Daniel J.

Committee members dc:contributor.committeemember

• Williams, Ryan K.
• Leonessa, Alexander

• reinforcement learning
• robotics
• humanoid robotics

Statement dc:rights

• In Copyright

Licence dc:rights.uri

• http://rightsstatements.org/vocab/InC/1.0/

Language dc:language.iso

en

Dc Identifier Other

vt_gsexam:44175

OAI identifier oai:identifier

oai:vtechworks.lib.vt.edu:10919/135528