STEADY: Simultaneous State Estimation and Dynamics Learning from Indirect Observations

TL;DR

STEADY introduces a method for learning neural stochastic kinodynamic models from noisy, indirect observations by combining state estimation and dynamics learning in an iterative EM framework, improving accuracy and robustness.

Contribution

The paper presents a novel technique that simultaneously estimates states and learns dynamics from indirect, noisy data using an EM approach with particle filtering and stochastic gradient updates.

Findings

Achieves higher accuracy than baseline methods.

Demonstrates robustness to observation noise.

Effective on both simulation and real-world data.

Abstract

Accurate kinodynamic models play a crucial role in many robotics applications such as off-road navigation and high-speed driving. Many state-of-the-art approaches in learning stochastic kinodynamic models, however, require precise measurements of robot states as labeled input/output examples, which can be hard to obtain in outdoor settings due to limited sensor capabilities and the absence of ground truth. In this work, we propose a new technique for learning neural stochastic kinodynamic models from noisy and indirect observations by performing simultaneous state estimation and dynamics learning. The proposed technique iteratively improves the kinodynamic model in an expectation-maximization loop, where the E Step samples posterior state trajectories using particle filtering, and the M Step updates the dynamics to be more consistent with the sampled trajectories via stochastic gradient ascent. We evaluate our approach on both simulation and real-world benchmarks and compare it with several baseline techniques. Our approach not only achieves significantly higher accuracy but is also more robust to observation noise, thereby showing promise for boosting the performance of many other robotics applications.