GUARD: Sensitivity-aware Uncertainty Attribution for Safe Traditional and Learning-based Robot Control
Overview
This paper presents GUARD, a framework that combines probabilistic state estimation with uncertainty identification to establish critical thresholds for robot control. By integrating probabilistic kernel optimization with sensitivity-aware uncertainty attribution, the system forwards estimated uncertainties to controllers to adapt safety margins and initiate safe backup trajectories. Experiments on traditional model predictive control (MPC) and reinforcement learning (RL) controllers show that GUARD reduces collision rates by 44-67% in RL evaluations and decreases backup policy activation by 50-75% under medium-to-high uncertainty profiles, while successfully avoiding collisions in MPC simulations.
Generated from attributed reports · Updated 2 days ago
Event evidence and corrections
0 attributed source owners. Ownership does not establish independent confirmation. Quantities are reported separately and are never added together.
Environment: simulation
“a reinforcement learning (RL) based controller. The results demonstrate that GUARD reduces collision rates by 44-67% in the RL-based evaluation, while decreasing backup policy activation by 50-75% in the medium- and high-uncertainty profiles. In the MPC simulation, GUARD successfully avoids collisions. By unifying the probabilistic kernel optimization algorithm from robot state estimation with the uncertainty estima”
Exact source · revision 1Source owner not reported
Report timeline
Follow attributed reports and material updates.
- arXiv Robotics — research abstractsSignalGUARD: Sensitivity-aware Uncertainty Attribution for Safe Traditional and Learning-based Robot Control
This paper presents GUARD, a framework that combines probabilistic state estimation with uncertainty identification to establish critical thresholds for robot control. By integrating probabilistic kernel optimization with sensitivity-aware uncertainty attribution, the system forwards estimated uncertainties to controllers to adapt safety margins and initiate safe backup trajectories. Experiments on traditional model predictive control (MPC) and reinforcement learning (RL) controllers show that GUARD reduces collision rates by 44-67% in RL evaluations and decreases backup policy activation by 50-75% under medium-to-high uncertainty profiles, while successfully avoiding collisions in MPC simulations.
Timezone · UTC
Article dates follow your selected timezone. Briefing editions use Hong Kong time (UTC+8).