2019 Wagner Prize Winner Reprise – AI at Work
You may never have heard of DIDI, but they are a ride sharing company like Uber and Lyft operating around the world. They provide on the order of 10 Billion rides per year. One of the core algorithms used in this industry is the matching algorithm that assigns drivers to riders. A simple solution is to process a ride request by assigning the “nearest” driver. In this talk, you will hear how a team from DIDI Research America, formulated a better algorithm that improves driver income, is “fair” to the driver pool, and yet maintains a high satisfaction rate for riders. In order to tackle the stochastic nature of demand and supply in the ride-hailing marketplace, they created a Semi-Markov Decision Process (MDP) Model and used Reinforcement Learning to solve this complicated model.
Come here about this fascinating work today at 3:40 pm in CC Room 6A.