Handling Missing Observations with an RNN-based Prediction-Update Cycle

TL;DR

This paper presents an RNN-based method inspired by Kalman filtering that effectively handles missing observations in time-series data, improving motion state estimation in tracking scenarios.

Contribution

It introduces a novel RNN architecture that incorporates masking and belief updates to manage missing data and outliers within a full temporal filtering cycle.

Findings

Successfully handles missing data in synthetic pedestrian tracking scenarios

Outperforms traditional methods in dealing with missing observations

Provides a unified RNN-based filtering approach

Abstract

In tasks such as tracking, time-series data inevitably carry missing observations. While traditional tracking approaches can handle missing observations, recurrent neural networks (RNNs) are designed to receive input data in every step. Furthermore, current solutions for RNNs, like omitting the missing data or data imputation, are not sufficient to account for the resulting increased uncertainty. Towards this end, this paper introduces an RNN-based approach that provides a full temporal filtering cycle for motion state estimation. The Kalman filter inspired approach, enables to deal with missing observations and outliers. For providing a full temporal filtering cycle, a basic RNN is extended to take observations and the associated belief about its accuracy into account for updating the current state. An RNN prediction model, which generates a parametrized distribution to capture the predicted states, is combined with an RNN update model, which relies on the prediction model output and the current observation. By providing the model with masking information, binary-encoded missing events, the model can overcome limitations of standard techniques for dealing with missing input values. The model abilities are demonstrated on synthetic data reflecting prototypical pedestrian tracking scenarios.