notesum.ai

Published at December 3

Cooperative Cruising: Reinforcement Learning based Time-Headway Control for Increased Traffic Efficiency

cs.MA

cs.AI

cs.LG

cs.SY

eess.SY

Released Date: December 3, 2024

Authors: Yaron Veksler¹, Sharon Hornstein, Han Wang, Maria Laura Delle Monache, Daniel Urieli

Aff.: ¹General Motors R&D Labs

Arxiv: http://arxiv.org/pdf/2412.02520v1

Refer to caption

SUMO parameters
step-length	0.5 seconds
lateral-resolution	0.4 meters
extrapolate-departpos	True
tau (default time-headway)	1.5 seconds
lcKeepRight	0
lcAssertive	3
lcSpeedGain	5
MDP parameters
$\frac{1}{C}$ (reward normalization)	1e-5
action range	[1.5, 6] seconds
num_control_segments	2
RLlib PPO parameters
num_rollout_workers	10
train_batch_size	2000
sgd_minibatch_size	128
clip_param	0.3
num_sgd_iter	30
use_gae	True
lambda	1
vf_loss_coeff	1
kl_coeff	0.2
entropy_coeff	0
learning_rate	5e-5