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

Environmental disturbances, such as sensor data noises, various lightingconditions, challenging weathers and external adversarial perturbations, areinevitable in real self-driving applications. Existing researches and testingshave shown that they can severely influence the vehicles perception ability andperformance, one of the main issue is the false positive detection, i.e., theghost object which is not real existed or occurs in the wrong position (such asa non-existent vehicle). Traditional navigation methods tend to avoid everydetected objects for safety, however, avoiding a ghost object may lead thevehicle into a even more dangerous situation, such as a sudden break on thehighway. Considering the various disturbance types, it is difficult to addressthis issue at the perceptual aspect. A potential solution is to detect theghost through relation learning among the whole scenario and develop anintegrated end-to-end navigation system. Our underlying logic is that thebehavior of all vehicles in the scene is influenced by their neighbors, andnormal vehicles behave in a logical way, while ghost vehicles do not. Bylearning the spatio-temporal relation among surrounding vehicles, aninformation reliability representation is learned for each detected vehicle andthen a robot navigation network is developed. In contrast to existing works, weencourage the network to learn how to represent the reliability and how toaggregate all the information with uncertainties by itself, thus increasing theefficiency and generalizability. To the best of the authors knowledge, thispaper provides the first work on using graph relation learning to achieveend-to-end robust navigation in the presence of ghost vehicles. Simulationresults in the CARLA platform demonstrate the feasibility and effectiveness ofthe proposed method in various scenarios.

Conquering Ghosts: Relation Learning for Information Reliability Representation and End-to-End Robust Navigation