A custom-reward PPO agent trained in a Unity 3D traffic simulation reduces serious collisions by 75% and vehicle-vehicle collisions by 79% relative to a baseline, while increasing total distance traveled by 345%.
(2019, March 7)
1 Pith paper cite this work. Polarity classification is still indexing.
1
Pith paper citing it
fields
cs.LG 1years
2025 1verdicts
REJECT 1representative citing papers
citing papers explorer
-
Safety-Prioritized, Reinforcement Learning-Enabled Traffic Flow Optimization in a 3D City-Wide Simulation Environment
A custom-reward PPO agent trained in a Unity 3D traffic simulation reduces serious collisions by 75% and vehicle-vehicle collisions by 79% relative to a baseline, while increasing total distance traveled by 345%.