Particle Filtering with Generative Trajectory Models for State Estimation and Control
Particle Filtering with Generative Trajectory Models for State Estimation and Control
This research paper introduces a method that combines particle filtering with generative trajectory models to enhance state estimation and control in robotics. The authors propose using a single pretrained diffusion model to generate competing hypotheses about state histories, along with predicted observations and action proposals. By reweighting particles based on how well their predicted observations match new measurements, the system improves both state estimation accuracy and task performance. Simulation experiments in partially observed navigation and vision-based manipulation show that this approach outperforms standard action generation methods using the same pretrained model without requiring additional test-time information or model modifications.
This preprint proposes using a pretrained diffusion model within a particle filtering framework to jointly perform state estimation and action selection, demonstrating improved performance in simulation for navigation and manipulation tasks without modifying the model weights.
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
- paper: not_reported
- code: not_reported
- weights: not_reported
- dataset: not_reported
Source excerpts and review record
Automatically extracted; no manual editorial approval recorded.
according to how well their predicted observations match the measurements. While particle filtering conventionally uses these weights only for state estimation, we also use them to guide action selection, favoring proposals associated with hypotheses that better explain the observations. Simulation experiments in partially observed navigation and vision-based manipulation demonstrate improvements in both state estim
Open source S4
arXiv:2611.11732v1 Announce Type: new Abstract: Generative models can support both state estimation and control in robotics. For control, a common approach is to use the model as a policy that directly samples actions conditioned on observation histories. However, models that jointly represent states, actions, and observations can also predict how the robot state and its environment evolve. We investigate whether the
Open source S2
Source：arXiv Robotics — research abstracts · arxiv.org