Uploaded January 2026 | Updated September 2026, 2 weeks ago
This Tech Talk covers different approaches for using reinforcement learning (RL) to develop and deploy control policies on real hardware. Instead of focusing on algorithm details, it compares strategies for training and running policies, emphasizing tradeoffs in hardware safety, training time, and system performance. The demo uses a Quanser Qube Servo 2 rotary pendulum, controlled by a policy running on a Raspberry Pi®, with training performed in MATLAB® and Simulink® on a PC. Key concepts include offline RL, training on hardware versus simulation, and the importance of validating policies before deployment. The talk also highlights challenges such as communication latency, computational limits of embedded processors, and the sim2real gap when transferring policies from simulation to physical systems.
Learn more:
- Reinforcement Learning With Hardware: Train Policy Deployed on Raspberry Pi: bit.ly/4a841gE
- Train Reinforcement Learning Agents to Control Quanser QUBE Pendulum: bit.ly/4iLt0se
Chapters:
00:00 Introduction
00:45 Setting up the problem
03:31 Different approaches to training a policy on hardware
06:37 Show demonstration of training on hardware
11:08 Training in a simulated environment
13:47 Run trained policy on real hardware
15:22 Reduce the sim2real gap
16:50 Closing thoughts
--------------------------------------------------------------------------------------------------------
Get a free product trial: goo.gl/ZHFb5u
Learn more about MATLAB: goo.gl/8QV7ZZ
Learn more about Simulink: goo.gl/nqnbLe
See what's new in MATLAB and Simulink: goo.gl/pgGtod
© 2025 The MathWorks, Inc. MATLAB and Simulink are registered trademarks of The MathWorks, Inc.
See mathworks.com/trademarks for a list of additional trademarks. Other product or brand names may be trademarks or registered trademarks of their respective holders.
This Tech Talk covers different approaches for using reinforcement learning (RL) to develop and deploy control policies on real hardware. Instead of focusing on algorithm details, it compares strategies for training and running policies, emphasizing tradeoffs in hardware safety, training time, and system performance. The demo uses a Quanser Qube Servo 2 rotary pendulum, controlled by a policy running on a Raspberry Pi®, with training performed in MATLAB® and Simulink® on a PC. Key concepts include offline RL, training on hardware versus simulation, and the importance of validating policies before deployment. The talk also highlights challenges such as communication latency, computational limits of embedded processors, and the sim2real gap when transferring policies from simulation to physical systems.
Learn more:
- Reinforcement Learning With Hardware: Train Policy Deployed on Raspberry Pi: bit.ly/4a841gE
- Train Reinforcement Learning Agents to Control Quanser QUBE Pendulum: bit.ly/4iLt0se
Chapters:
00:00 Introduction
00:45 Setting up the problem
03:31 Different approaches to training a policy on hardware
06:37 Show demonstration of training on hardware
11:08 Training in a simulated environment
13:47 Run trained policy on real hardware
15:22 Reduce the sim2real gap
16:50 Closing thoughts
--------------------------------------------------------------------------------------------------------
Get a free product trial: goo.gl/ZHFb5u
Learn more about MATLAB: goo.gl/8QV7ZZ
Learn more about Simulink: goo.gl/nqnbLe
See what's new in MATLAB and Simulink: goo.gl/pgGtod
© 2025 The MathWorks, Inc. MATLAB and Simulink are registered trademarks of The MathWorks, Inc.
See mathworks.com/trademarks for a list of additional trademarks. Other product or brand names may be trademarks or registered trademarks of their respective holders.
