Application scenarios
Power generation control
Generators adjust their output to balance supply and demand at minimum cost as wind power fluctuates.
Open scenario →Cloud job scheduling
Dispatchers assign arriving jobs to servers to reduce job waiting time and energy consumption.
Open scenario →Distribution voltage control
Rooftop solar inverters adjust their reactive power to keep every bus voltage within limits as solar output rises and falls.
Open scenario →Robotaxi fleet rebalancing
City districts send their idle robotaxis ahead of demand so that riders find a car as travel patterns shift across the city.
Open scenario →More scenarios in progress
We are building further scenarios, such as traffic signal control.
Related work
How to use analytical models to train AI agents in large-scale complex systems
Incorporating prior knowledge of the system into learning to scale to large multi-agent systems.
Related work
Routing for connected vehicles
Exploiting homogeneity among vehicles to scale routing to large fleets via mean-field reinforcement learning.