The main tool used to conduct experiments is SUMO-RL framework.
Ensemble methods used:
- Majority Voting
- Soft Voting
- Transformed Rank Voting
- Average Voting - Boltzmann Probs
Reward is defined as the change of the cumulative vehicle delay
Action: Choose the next Green phase
State Representation: Vector of dimension
Environment: Simulation of Urban MObility (SUMO)
A single intersection consisted of:
- 2 Incoming - 2 Outgoing approaches
- Totally 8 Lanes
- 8 Permitted Movement Signals
- Sythetic Data built on SUMO:
- Approach Length:
$300m$ - Cycle Traffic Plan Duration:
$82s$
- Approach Length: