Denoising Particle Filters: Learning State Estimation with Single-Step Objectives
This addresses the issue of interpretability and training cost in learning-based state estimation for robotics, offering a composable alternative to end-to-end methods.
The paper tackles the problem of state estimation in robotics by proposing a particle filtering algorithm where models are trained from individual state transitions using a denoising score matching objective, achieving competitive performance compared to end-to-end trained baselines in simulation.
Learning-based methods commonly treat state estimation in robotics as a sequence modeling problem. While this paradigm can be effective at maximizing end-to-end performance, models are often difficult to interpret and expensive to train, since training requires unrolling sequences of predictions in time. As an alternative to end-to-end trained state estimation, we propose a novel particle filtering algorithm in which models are trained from individual state transitions, fully exploiting the Markov property in robotic systems. In this framework, measurement models are learned implicitly by minimizing a denoising score matching objective. At inference, the learned denoiser is used alongside a (learned) dynamics model to approximately solve the Bayesian filtering equation at each time step, effectively guiding predicted states toward the data manifold informed by measurements. We evaluate the proposed method on challenging robotic state estimation tasks in simulation, demonstrating competitive performance compared to tuned end-to-end trained baselines. Importantly, our method offers the desirable composability of classical filtering algorithms, allowing prior information and external sensor models to be incorporated without retraining.