Skip to content
arXiv Robotics — research abstracts· Luis Pabon, Johannes K\"{o}hler, Revant Kasichainula, Marco Pavone·· 4 days agoEditorial score65

Full-Order Safety for Nonlinear Reduced-Order MPC

Full-Order Safety for Nonlinear Reduced-Order MPC

Summary

The paper presents CoRE-MPC, a robust model predictive control approach that uses reduced-order models to ensure safety and constraint satisfaction for high-dimensional nonlinear systems. By incorporating data-driven error radii, the method guarantees recursive feasibility and provides error bounds. Validated on complex simulations, including a 9,768-state soft robot, CoRE-MPC achieves 13 to 61 times faster performance than full-order robust MPC with only a 27 to 35% increase in closed-loop cost.

Source: arXiv Robotics — research abstracts · Read original article ↗

Loading article text…

What the source reports

Publisher-reported claims, with original evidence. These results have not been independently verified by RoboSignal.

Environment
SimulationOpen source S4

Reported numbers

  • state_count

    80

    View original evidence
    ach is validated in simulation on an 80-state oscillator chain, a Burgers equation, and a 9,768-state finite-element soft robot, where a two-state ROM with data-driven bounds keeps all 1,628 nodes clear of a wall at a median solve time below 1 ms for a 20 ms control period. Across these examples, CoRE-MPC is 13 to 61 times faster than full-order robust MPC at only 27 to 35% higher closed-loop cost, whereas existing c
    Open source S4

  • state_count

    9,768

    View original evidence
    ach is validated in simulation on an 80-state oscillator chain, a Burgers equation, and a 9,768-state finite-element soft robot, where a two-state ROM with data-driven bounds keeps all 1,628 nodes clear of a wall at a median solve time below 1 ms for a 20 ms control period. Across these examples, CoRE-MPC is 13 to 61 times faster than full-order robust MPC at only 27 to 35% higher closed-loop cost, whereas existing c
    Open source S4

  • node_count

    1,628

    View original evidence
    ach is validated in simulation on an 80-state oscillator chain, a Burgers equation, and a 9,768-state finite-element soft robot, where a two-state ROM with data-driven bounds keeps all 1,628 nodes clear of a wall at a median solve time below 1 ms for a 20 ms control period. Across these examples, CoRE-MPC is 13 to 61 times faster than full-order robust MPC at only 27 to 35% higher closed-loop cost, whereas existing c
    Open source S4

  • solve_time

    <1

    View original evidence
    ach is validated in simulation on an 80-state oscillator chain, a Burgers equation, and a 9,768-state finite-element soft robot, where a two-state ROM with data-driven bounds keeps all 1,628 nodes clear of a wall at a median solve time below 1 ms for a 20 ms control period. Across these examples, CoRE-MPC is 13 to 61 times faster than full-order robust MPC at only 27 to 35% higher closed-loop cost, whereas existing c
    Open source S4

  • control_period

    20

    View original evidence
    ach is validated in simulation on an 80-state oscillator chain, a Burgers equation, and a 9,768-state finite-element soft robot, where a two-state ROM with data-driven bounds keeps all 1,628 nodes clear of a wall at a median solve time below 1 ms for a 20 ms control period. Across these examples, CoRE-MPC is 13 to 61 times faster than full-order robust MPC at only 27 to 35% higher closed-loop cost, whereas existing c
    Open source S4

  • speed_up_factor

    13–61

    View original evidence
    ach is validated in simulation on an 80-state oscillator chain, a Burgers equation, and a 9,768-state finite-element soft robot, where a two-state ROM with data-driven bounds keeps all 1,628 nodes clear of a wall at a median solve time below 1 ms for a 20 ms control period. Across these examples, CoRE-MPC is 13 to 61 times faster than full-order robust MPC at only 27 to 35% higher closed-loop cost, whereas existing c
    Open source S4

  • closed_loop_cost_increase

    27–35

    View original evidence
    ach is validated in simulation on an 80-state oscillator chain, a Burgers equation, and a 9,768-state finite-element soft robot, where a two-state ROM with data-driven bounds keeps all 1,628 nodes clear of a wall at a median solve time below 1 ms for a 20 ms control period. Across these examples, CoRE-MPC is 13 to 61 times faster than full-order robust MPC at only 27 to 35% higher closed-loop cost, whereas existing c
    Open source S4
Source excerpts and review record

Automatically extracted; no manual editorial approval recorded.

ach is validated in simulation on an 80-state oscillator chain, a Burgers equation, and a 9,768-state finite-element soft robot, where a two-state ROM with data-driven bounds keeps all 1,628 nodes clear of a wall at a median solve time below 1 ms for a 20 ms control period. Across these examples, CoRE-MPC is 13 to 61 times faster than full-order robust MPC at only 27 to 35% higher closed-loop cost, whereas existing c

Open source S4

ertified schemes are either infeasible or at least twice as conservative. Code: [released upon publication].

Open source S5

Source：arXiv Robotics — research abstracts · arxiv.org

#Research#Reinforcement learning#Sim-to-real#Evaluation
Timezone · UTC

Article dates follow your selected timezone. Briefing editions use Hong Kong time (UTC+8).