python3 -m sample --pg-type td3

check out flow/flow/envs for detailed information

• one action for each intersection
• range within 0 and 1

• fully observable state space (TrafficLightGridEnv)
  • for vehicles:
    1. velocity
    2. distance from the next intersection
    3. the unique edge it is traveling on
  • for each traffic light:
    1. current state (the flowing direction)
    2. last changed time
    3. whether it's yellow

• large delay penalty
• switch penalty

01 23 45 11 10

9 8

7 6

• step function is at flow/envs/base.py

• env.k.vehicle.get_speed(veh_id)
• env.k.network.max_speed()
• env.network.node_mapping: iteratively give node and edge
• env.k.vehicle.get_edge(veh_id): current edge veh_id is at
• env.k.vehicle.get_ids_by_edge(edge): gives ids
  • for k in env.k.network.get_edge_list()