←
Thibaut Vidal735 posts
Thibaut Vidal avatar

Thibaut Vidal

@vidalthi

Professor & SCALE-AI Chair at MAGI @polymtl | @ivadolabs advisor Tweets on #ORMS & #Trustworthy #MachineLearning Open-source codes: github.com/vidalt

◷ Joined May 2017▧ 149 Photos and videos
Thibaut Vidal avatar
Thibaut Vidal@vidalthi·Nov 9

Great algorithms go much further when everyone can build on them. I'm very excited about this upcoming CVRPLib challenge in 2027 & seeing @tigfoundation's algorithms play a key role under an open data license!

The TIG thesis in action. Eduardo Queiroga (UFBA) and Rafael Martinelli (PUC-Rio), both of whom help organise CVRPLib's vehicle-routing challenges, have built on Thibaut Vidal's HGS Advance algorithm, developed for our Vehicle Routing Challenge. Their modified implementation is being used to generate best-known solutions for CVRPLib's upcoming Vehicle Routing Problem with Time Windows (VRPTW) Challenge. These are the solutions competitors will aim to beat. Their work has also been published under the TIG Open Data License v3.0. This is exactly what TIG was designed to make possible. We want to create an environment where researchers can build on each other's contributions, rather than having every finding exist in isolation. An algorithm developed for one challenge can become the basis for another, giving people a foundation to test new ideas, push its limits and make further discoveries. This is how research can become cumulative. Less duplicated effort, more collaboration and faster progress.  We're not just rewarding individual achievements. We're building a network where those breakthroughs can lead to many more, and then some. More on the work - Implementation: github.com/Galgos-PUC-Rio/hg… Competition code to paper: galgos.inf.puc-rio.br/cvrpli… VRPTW Challenge overview: galgos.inf.puc-rio.br/cvrpli…

Post attachment
Thibaut Vidal avatar
Thibaut Vidal@vidalthi·Jul 13

The debris removal challenge is to jointly decide: - which debris each spacecraft should target - in which order & at what time - using which orbital transfers These decisions are coupled, as debris motion and orbital-plane drift make transfer costs strongly time-dependent.

Thibaut Vidal avatar
Thibaut Vidal@vidalthi·Jul 13

As satellite populations grow, conjunction and collision risks rise. A collision can create fragments that cause further collisions... a "collisional cascade" (Kessler syndrome), potentially making valuable orbital regions unsafe. Active debris removal is becoming essential.

Post attachment
Thibaut Vidal avatar
Thibaut Vidal@vidalthi·Jul 13

How can we plan orbital debris removal at scale, with multiple spacecraft operating simultaneously? 🚀 We present our work today at #IFORS2026: 🕑 14:15–15:45 📍 JUR – Hörsaal 15 📋 ME-33 — Space Logistics

0:21
Thibaut Vidal avatar
Thibaut Vidal@vidalthi·Jul 10

Co-authored with @FerryJulien12, it concludes nearly three years of manual curation, reading, and synthesis. We are submitting now and warmly welcome feedback! I will also be attending #IFORS2026 and am very pleased to discuss this topic!

Thibaut Vidal avatar
Thibaut Vidal@vidalthi·Jul 10

This matters even more as predictive models are embedded in (increasingly autonomous) opt and decision pipelines. Models with similar accuracy might behave very differently downstream across key trustworthiness dimensions. CO helps select and audit good models.

Thibaut Vidal avatar
Thibaut Vidal@vidalthi·Jul 10

From decision trees and sparse models to ensembles and neural networks, #combinatorial #optimization methods are useful for training, selecting, explaining, auditing, verifying, and compressing models, with modular constraints, guarantees, and trade-off analysis.

Thibaut Vidal avatar
Thibaut Vidal@vidalthi·Jun 30

How much does routing optimization matter? Compared with Google OR-Tools and Hexaly, HGS-Advance shows substantial gains. In large supply chains, even a 4% reduction in travel distance translates into millions in savings.

Post attachment
Thibaut Vidal avatar
Thibaut Vidal@vidalthi·Jun 29

Finally, there is huge learning and educational value. Anyone, with or without an academic or CS background, can study a challenge, learn from open-source state-of-the-art methods, and propose improvements (small or large, assisted or not by agents). Congratulations! 👏

Thibaut Vidal avatar
Thibaut Vidal@vidalthi·Jun 29

#TIG’s game-changing innovation model has also been a key motivator. The code takes inspiration from successful ideas in previous contributions, notably `fast_lane_v4` 🙏, and will likewise be built upon by new contributors, sustaining a continuous stream of improvements🚣

Thibaut Vidal avatar
Thibaut Vidal@vidalthi·Jun 29

This open-source code in #Rust is highly scalable and built on decomposition and consensus strategies: github.com/tig-foundation/ti… It *is designed* for large instances with 1000s of delivery addresses. I did not initially think simple improvements of this magnitude were achievable.

Thibaut Vidal avatar
Thibaut Vidal@vidalthi·Jun 29

The votes have been cast: HGS "Advance" is now part of the TIG (@tigfoundation) ecosystem and adopted by benchmarkers! It is now, by a very significant margin, the SOTA for #VRPTW, substantially improving upon DIMACS challenge-winning results and existing commercial solvers.

Post attachment