news

Let us take you on an extraordinary journey during which we show how autonomous vehicles can disturb our cities by applying reinforcenment learning in their route choice. 🎤 This Sunday Nov 10 at 2:30 PM Anastasia and Onur are taking up the reins and waiting for you in Copernicus Science Centre. So don’t wait and register for our tutorial. Reach out for more info

🚗🚗🚗 We are proud to share our achievements in urban traffic dynamics at 17th European Workshop on Reinforcement Learning. Meet Onur and Anastasia who are proving that RL-enabled autonomous vehicles can optimize behavioral objectives by learning to choose better routes! Just take a look how we came to this conclusion 👈

Last Saturday we took part in Run for a Good Cause 🏃‍♂️ 🏃‍♀️ 🏃‍♂️ 🏃‍♀️ - a charity event organised by the Faculty of Mathematics and Computer Science to support Tadzio Kaminski, the son of our colleague. The competition has faded into the background, it was a great chance to help those in need and have fun while doing this. Our team composed of four did a good job!

Last week in Edmonton during 27th IEEE International Conference on Intelligent Transportation Systems Rafal presented our work on how we can use Reinforcement Learning to optimize platform market entry strategies. 🚕 Platforms like Uber, Bolt, Lyft or Didi may use it (based on our MoMaS platform growth simulator) to improve their strategies (fares, discounts and incentives). Check out paper abstract here

1st of October we signed ✒️ Michal Bujak and Farnoud Ghasemi for our SUM project. This is the first day of the official cooperation with them as PhD Students. We are sure they both will bring added value 💡💡💡 to the team.

This is a happy end of the super long story that started in 2020, with results and submission in 2021, followed with 2 rejections from journals, one rejection due to lack of reviewers and a lot of minor revisions. In this particular case we waited 3 years to publish our results, which can be now followed and developed in npj Sustainable Mobility and Transport. In this paper we proved that we can identify pooled rides of up to 14 people, with occupancies of 6 and more, with our hyper-pool algorithm. This while remaining attractive to all travellers (against private ride) and in many cases even competing with PT (in Amsterdam, where PT quality is excellent). Already 200 trip requests per 30-min batch (4000 trips per hour) in Amsterdam allows to identify high-occupancy hyper-pooled trips. 🔖 Our open access publication is available here!