Pierre Alliez: a pioneer in digital geometry processing

"During my post-doctoral studies at the University of Southern California, where I worked with Mathieu Desbrun, professor at the California Institute of Technology, " explains Pierre Alliez, "I began consolidating and identifying this topic, which wasn't a field of its own. At the time you either had computer graphics or algorithmic geometry. However, from 2003 a community began forming around the subject of digital geometry processing. When I later applied to Inria Sophia-Antipolis at the end of 2001, Jean-Daniel Boissonnat, leader of the PRISME team, welcomed my research project with open arms. This team became GEOMETRICA. Its aim is to develop an axiomatic approach to geometric computation. The vanguard of computer science is located at Inria. When I submitted my application for the ERC Grant, without any obligation in terms of results on their part, the Institute showed me the same enthusiasm and the same confidence, which is great, particularly when you like working independently like I do ."

The digitisation of geometry , according to Pierre Alliez and Mathieu Desbrun, consists of devising the equivalent of signal processing for 3D forms. For these two researchers, this is the logical follow-up to the digitisation of sound during the 1970s and 80s, then of images and lastly videos during the 2000s. However, digital geometry processing is related to the processing of increasingly disparate and uncertain data. "The first challenge of this research relates to the fact that we are faced with a technological paradox ," says the researcher, "we thought that the data would follow the evolution in sensors, but this isn't the case. The data require ever more processing and have never been as imperfect as they are today due to the diversification in acquisition methods and changing usages (super-resolution and new acquisition paradigms, including community data such as Flickr) ."

A data-processing time-saving of three weeks before simulation

The IRON solution  proposes robust and tolerant algorithms capable of withstanding the imperfection and diversity of any type of data, in contrast to the current methodology that consists of locating, converting or sorting data before processing. "It's here that we find the technological obstacle associated with IRON, our challenge number 2 ." The other challenge is societal in nature, a concept that Pierre Alliez calls the "new frontier" and on which he places a particular emphasis. "After the era of tailored items once reserved for the elite, followed by the era of mass production, I'm convinced that we're now moving into the mass-tailored era. This project won't change society but it might contribute to it ." These "tough" algorithms will need to stand up to digital geometry processing, which will have multiple applications. They will be used by a variety of engineers, including in the automotive and aeronautic fields, who employ computer-aided engineering and will improve in terms of efficiency.

Computational engineering  is taking the place of both the digital model of the physical prototype, and experience-based computing. Consequently, engineers are speeding up the design cycle by "testing the real deal" in order to improve their design and foresight capabilities. However, although simulation is routinely used, converting a final CAD model (from a production viewpoint) into a model ready for simulation requires three weeks of interactive trial-and-error processing for conversion, which considerably curbs the true potential of computational engineering. The initial goal of computational engineering was to bridge the gap between modelling and simulation. For example, the results of simulation - which takes 5,000 hours of parallel computing and one hour of actual time - might suggest raising a car bonnet by 5 cm and therefore going back to the design phase. However, because every return to the simulation phase takes three weeks, the design process is slowed down.

In concrete terms, IRON means that engineers save three weeks of data processing before simulation. This is the first step in mass-tailoring, which will also be applied to medicine, simulation (for sustainable cities, geology) and what is known as freeform architecture. It is also the promise linked with the ERC grant awarded to Pierre Alliez, who has given himself five years to achieve this.