Sampling-based Model Predictive Control as Sequential Monte Carlo Inference
Overview
This paper presents a unified Sequential Monte Carlo (SMC) inference algorithm that encompasses various sampling-based Model Predictive Control methods, including predictive sampling, MPPI, and the cross-entropy method. The authors evaluate this approach on MuJoCo non-linear control tasks, open-loop navigation through obstacle fields, 7-DoF motion planning, and non-prehensile push-T scenarios, both in simulation and on a Franka FR3 robot. Key results include a 30% reduction in collision rates at a 28-knot horizon compared to non-resampling baselines, and significant improvements in push-task success rates from 1% to 100% in simulation and from 10% to 90% on hardware when using programmable proposals.
Generated from attributed reports · Updated 11 hours 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
“nt axes: the proposal, the update rule that reduces cost-weighted particles to a nominal plan, and the effective-sample-size threshold for resampling. We evaluate SMC-based planning on MuJoCo non-linear control, open-loop navigation through obstacle fields, 7-DoF motion planning, and non-prehensile push-T, in simulation and on a Franka FR3. Resampling combats particle degeneracy at long horizons. At a 28-knot horizon”
Exact source · revision 1Source owner not reported
Report timeline
Follow attributed reports and material updates.
- arXiv Robotics — research abstractsSignalSampling-based Model Predictive Control as Sequential Monte Carlo Inference
This paper presents a unified Sequential Monte Carlo (SMC) inference algorithm that encompasses various sampling-based Model Predictive Control methods, including predictive sampling, MPPI, and the cross-entropy method. The authors evaluate this approach on MuJoCo non-linear control tasks, open-loop navigation through obstacle fields, 7-DoF motion planning, and non-prehensile push-T scenarios, both in simulation and on a Franka FR3 robot. Key results include a 30% reduction in collision rates at a 28-knot horizon compared to non-resampling baselines, and significant improvements in push-task success rates from 1% to 100% in simulation and from 10% to 90% on hardware when using programmable proposals.
Event coverage history
There is not enough continuous observation data to show a trend.
Timezone · UTC
Article dates follow your selected timezone. Briefing editions use Hong Kong time (UTC+8).