Sign in Subscribe
Artificial neural networks

Accelerated Neural MPC with Safety Guarantees: A Novel Framework for Robotics and Automation

Strategic Insights

— 1 min read

Research from Shanghai University proposes a novel framework called Barrier-integrated Adaptive Neural Model Predictive Control (BAN-MPC) that integrates neural networks' fast computation with Model Predictive Control's (MPC) constraint-handling capability. The framework is designed to ensure strict safety in robotics and automation systems while reducing online computational complexity. The study presents hardware-in-the-loop (HIL) experiments on Jetson Nano, demonstrating that BAN-MPC solves 200 times faster than traditional MPC, enabling collision-free navigation with control error below 5% under model parameter variations within 15%.

Key Takeaways:

  • The BAN-MPC framework synergizes neural networks' fast computation with MPC's constraint-handling capability, ensuring strict safety in robotics and automation systems.
  • The framework integrates an offline-learned neural value function into the optimization objective of a Short-horizon MPC, substantially reducing online computational complexity.
  • A second neural network is used to learn the sensitivity of the value function to system parameters, allowing adaptive adjustment of the neural value function based on model parameter changes.
  • The HIL experiments on Jetson Nano show that BAN-MPC solves 200 times faster than traditional MPC and enables collision-free navigation with control error below 5% under model parameter variations within 15%.
  • The research has been peer-reviewed and published in IEEE Robotics and Automation Letters.
  • The study proposes a novel control strategy that balances safety and speed in robotics and automation systems.
  • The framework has the potential to be applied in various robotics and automation applications where safety and speed are critical.

Statistics:

  • BAN-MPC solves 200 times faster than traditional MPC in HIL experiments on Jetson Nano.
  • The control error is below 5% under model parameter variations within 15% using BAN-MPC.
  • The framework reduces online computational complexity by integrating an offline-learned neural value function into the optimization objective of a Short-horizon MPC.
  • The HIL experiments demonstrate the effectiveness of BAN-MPC in enabling collision-free navigation in robotics and automation systems.

Sources:

  • Safety Meets Speed: Accelerated Neural Mpc With Safety Guarantees and No Retraining. Ieee Robotics and Automation Letters, 2025;10(11):11411-11418.
  • Ieee Robotics and Automation Letters can be contacted at: Ieee-inst Electrical Electronics Engineers Inc, 445 Hoes Lane, Piscataway, NJ 08855-4141, USA.
  • Our news journalists report that additional information may be obtained by contacting Liang Xu, Shanghai University, Sch Future Technol, Shanghai 200444, People's Republic of China.
  • Additional authors for this research include Kaikai Wang, Tianxun Li, Keyou You, and Qinglei Hu.

Read more