English (United Kingdom)

https://curated-unify.zendy.io/wp-json/zendy-region/v1/featured_content/oa?rat=en

https://curated-unify.zendy.io/wp-json/zendy-region/v1/highlighted_journal/

Global: Zendy Plus annual

Presents the access of premium content as premium feature

Premium Content

Presents the keyphrase highlighting as premium feature

Keyphrase Highlighting

Presents the summarisation as premium feature

Summarisation

Insights

Presents the pdf analysis as premium feature

PDF Analysis

Presents the zaia usage as premium feature

ZAIA

Global: Zendy Plus monthly

Global: Zendy Tools annual

Global: Zendy Tools monthly

Global: Zendy Free Plan

Autonomous vehicles and mobile robotic systems are typically equipped withmultiple sensors to provide redundancy. By integrating the observations fromdifferent sensors, these mobile agents are able to perceive the environment andestimate system states, e.g. locations and orientations. Although deep learningapproaches for multimodal odometry estimation and localization have gainedtraction, they rarely focus on the issue of robust sensor fusion - a necessaryconsideration to deal with noisy or incomplete sensor observations in the realworld. Moreover, current deep odometry models suffer from a lack ofinterpretability. To this extent, we propose SelectFusion, an end-to-endselective sensor fusion module which can be applied to useful pairs of sensormodalities such as monocular images and inertial measurements, depth images andLIDAR point clouds. Our model is a uniform framework that is not restricted tospecific modality or task. During prediction, the network is able to assess thereliability of the latent features from different sensor modalities andestimate trajectory both at scale and global pose. In particular, we proposetwo fusion modules - a deterministic soft fusion and a stochastic hard fusion,and offer a comprehensive study of the new strategies compared to trivialdirect fusion. We extensively evaluate all fusion strategies in both publicdatasets and on progressively degraded datasets that present syntheticocclusions, noisy and missing data and time misalignment between sensors, andwe investigate the effectiveness of the different fusion strategies inattending the most reliable features, which in itself, provides insights intothe operation of the various models.

Learning Selective Sensor Fusion for States Estimation