Overall Objectives
Research Program
Application Domains
Software and Platforms
New Results
Bilateral Contracts and Grants with Industry
Partnerships and Cooperations
XML PDF e-pub
PDF e-Pub

Section: New Results

Other applications

Real-time Reaction-Diffusion Simulation: a Machine Learning Technique

Participants : Thomas Degris, Nejib Zemzemi.

Carmen is an Inria team working on modeling the electrical activity of the human heart. Their models are mainly based on reaction-diffusion equations. These methods are expansive in terms of computational costs which limits their use in practice. More specifically, some recent chirurgical intervention techniques on the heart (atrial ablation) requires to identify the source of the electrical wave. Finding such sources requires an optimization procedure. Using classical methods, this procedure is very heavy computationally.

In this project, our goal is to reduce the computational cost using supervised learning techniques. The idea is to replace the incremental resolution of partial differential equations by more suitable data structures for real-time running. Starting from data generated by simulating different excitations scenari on a human atria, this data is afterwords used as a training data set for machine learning algorithms. This approach will allow a faster optimization procedure.

This work is in collaboration with Nejib Zemzemi from the Inria Carmen team. This project is in preliminary steps.

Appearance-based segmentation of indoors/outdoors sequences of spherical views

Participant : David Filliat.

In collaboration with Patrick Rives and Alexandre Chapoulie from the Arobas team at Inria Sophia-Antipolis, we developped a method for environment segmentation based on spherical views [41] . Navigating in large scale, complex and dynamic environments requires reliable representations able to capture metric, topological and semantic aspects of the scene for supporting path planing and real time motion control. In a previous work, we addressed metric and topological representations thanks to a multi-cameras system which allows building of dense visual maps of large scale 3D environments. The map is a set of locally accurate spherical panoramas related by 6dof poses graph. The work presented here is a further step toward a semantic representation. We aim at detecting the changes in the structural properties of the scene during navigation. Structural properties are estimated online using a global descriptor relying on spherical harmonics which are particularly well-fitted to capture properties in spherical views. A change-point detection algorithm based on a statistical Neyman-Pearson test allows us to find optimal transitions between topological places. Results are presented and discussed both for indoors and outdoors experiments.

Modelling Stop Intersection Approaches using Gaussian Processes

Participant : David Filliat.

In collaboration with Javier-Ibanez Guzman and Alexandre Armand from Renault, we developped an approach toward the development of an electronic co-pilot adapted to the driver behavior [39] . Indeed, each driver reacts differently to the same traffic conditions, however, most Advanced Driving Assistant Systems (ADAS) assume that all drivers are the same. This work proposes a method to learn and to model the velocity profile that the driver follows as the vehicle decelerates towards a stop intersection. Gaussian Processes (GP), a machine learning method for non-linear regressions are used to model the velocity profiles. It is shown that GP are well adapted for such an application, using data recorded in real traffic conditions. GP allow the generation of a normally distributed speed, given a position on the road. By comparison with generic velocity profiles, benefits of using individual driver patterns for ADAS issues are presented.