Project Description

DVS-Straight is an international collaboration project funded by the ITH department of AIST. In this project, CNRS-AIST JRL is partnering with UTBM, CIAD lab (France) and UPJV, MIS lab (France) with the aim of mastering the robot trajectories during image-based visual servoing directly using the dense brightness of captured images as input of the control law (Direct Visual Servoing: DVS). Indeed, DVS is first known for its submillimetric precision and recently for a quite large convergence domain, e.g. with DDVS. However, trajectories are not always satisfactory.

The goals are:

  • To develop an international network of labs and researchers to work on the DVS-Straight topic
  • To exploit better the control theory for improving the conditioning
  • To encapsulate DVS in multi-objective control frameworks for condisering explicit trajectory constraints
  • To research the optimal field-of-view maximizing the motion perceptibility
all for better behavior of DVS-based control laws and hence applicability in a much larger variety of scenarios and contexts.

A series of international workshops will be organized to share the research, the results and the open-source software as well as to develop the research community on these topics.

Title Authors Conference/Book Year bib pdf
Photometric Visual Gyroscope for Full-View Spherical Camera A. André, G. Caron IEEE/CVF Conference on Computer Vision and Pattern Recognition Workshops 2022
Direct Visual Servoing in the non-linear scale space of camera pose G. Caron, Y. Yoshiyasu IAPR/IEEE International Conference on Pattern Recognition 2022