Indeed, I do favor MPC over RL.
I've been making that point since at least 2016.
RL requires ridiculously large numbers of trials to learn any new task.
In contrast MPC is zero shot: If you have a good world model and a good task objective, MPC can solve new tasks without any task-specific learning.
That's the magic of planning.
It doesn't mean that RL is useless, but its use should be a last resort.
📢 Für alle Kurzentschlossenen: Das Open Data Beer findet diesen Donnerstag in Basel statt! 🍻
📅 19. September 2024
Feiere mit uns das 5-jährige Jubiläum des Datenportals Basel-Stadt und tausche dich beim Apéro aus.
https://t.co/oMd1sw0OU9
Awesome workshop by @SATW_ch at @UZH_ch#bridgelab regarding innovation power and what needs to be improved @swissinnova
Full study and recommendations: https://t.co/qIx5xoOlP3
Important exhibit at @EPFL : DigitalDilemma by @ICRC in collaboration with also @ETH_en
If you can’t visit the campus, take the digital tour: https://t.co/C3sDMbonr9
Sneak preview: our recent concept of a **dynamic virtual power plant** (DVPP) will be hard-coded in the new European grid code.
Learn more about it here: https://t.co/ygo68DqzHs
@AGISTINProject@posytyf @nccr_automation @ChamperyPower
@energygovuk are there some interactive visualisations available for "Monthly deployment of all solar photovoltaic capacity in the United Kingdom" dataset? Are the historic past monthly datasets available as well? I only found the recent data here: https://t.co/YoBAqDJnHN
Within the @nccr_automation , we have opened 25 collaborative research positions: https://t.co/5de8y97y67
... plus another dozen vacant positions with individual PIs. If you are interested, then contact your favorite PIs directly: https://t.co/OEyIEdoaTl