Context Transfer in Reinforcement Learning Using Action-Value Functions

Figure 5

Crossroad Traffic Controller. Old system: Sensors: distance sensor, , : distance to first car in vertical lane, : distance to first car in horizontal lane, : vertical lane is green, : horizontal lane is green, , : change the vertical lane to green, : change the horizontal lane to green, : no action. New system: Sensors: camera, , : cars’ existence coding in the first ten squares of the vertical lane, : cars’ existence coding in the first ten squares of the horizontal lane, : vertical lane is green, : horizontal lane is green, , : change the light, : no action.