# Dynamic Traffic Signal Optimization Configuration experiment: name: "traffic_rl_mtech" version: "1.0" description: "M.Tech Dynamic Traffic Signal Optimization using Deep RL" environment: simulation_time: 3600 # 1 hour simulation step_size: 1 # SUMO step size in seconds yellow_time: 3 min_green_time: 10 max_green_time: 60 warmup_time: 300 # 5 minutes warmup network: type: "single_intersection" lanes_per_direction: 2 max_speed: 50 # km/h intersection_size: 50 # meters agent: algorithm: "D3QN" # Dueling Double DQN state_size: 20 action_size: 8 learning_rate: 0.0001 gamma: 0.95 epsilon_start: 1.0 epsilon_end: 0.01 epsilon_decay: 0.995 memory_size: 100000 batch_size: 64 target_update_freq: 100 hidden_layers: [256, 128, 64] training: episodes: 2000 max_steps_per_episode: 1000 save_freq: 100 eval_freq: 50 log_freq: 10 evaluation: test_episodes: 10 baseline_methods: ["fixed_time", "actuated", "random"] metrics: ["delay", "queue_length", "throughput", "emissions", "fuel"] paths: models: "models/" data: "data/" logs: "logs/" results: "results/" sumo_configs: "sumo_configs/"