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 navigation consists in an agent being able to navigate withouthuman intervention or supervision, it affects both high level planning and lowlevel control. Navigation is at the crossroad of multiple disciplines, itcombines notions of computer vision, robotics and control. This work aimed atcreating, in a simulation, a navigation pipeline whose transfer to the realworld could be done with as few efforts as possible. Given the limited time andthe wide range of problematic to be tackled, absolute navigation performanceswhile important was not the main objective. The emphasis was rather put onstudying the sim2real gap which is one the major bottlenecks of modern roboticsand autonomous navigation. To design the navigation pipeline four mainchallenges arise; environment, localization, navigation and planning. TheiGibson simulator is picked for its photo-realistic textures and physicsengine. A topological approach to tackle space representation was picked overmetric approaches because they generalize better to new environments and areless sensitive to change of conditions. The navigation pipeline is decomposedas a localization module, a planning module and a local navigation module.These modules utilize three different networks, an image representationextractor, a passage detector and a local policy. The laters are trained onspecifically tailored tasks with some associated datasets created for thosespecific tasks. Localization is the ability for the agent to localize itselfagainst a specific space representation. It must be reliable, repeatable androbust to a wide variety of transformations. Localization is tackled as animage retrieval task using a deep neural network trained on an auxiliary taskas a feature descriptor extractor. The local policy is trained with behavioralcloning from expert trajectories gathered with ROS navigation stack.

Learning Synthetic to Real Transfer for Localization and Navigational Tasks