Deep reinforcement learning for traffic signal control with consistent state and reward design approach
Why this work is in the frame
A frame that forgets how it found something cannot be audited. These are the routes that admitted this work.
Bibliographic record
Abstract
Intelligent Transportation Systems are essential due to the increased number of traffic congestion problems and challenges nowadays. Traffic Signal Control (TSC) plays a critical role in optimizing the traffic flow and mitigating the congestion within the urban areas. Various research works have been conducted to enhance the behavior of TSCs at intersections and subsequently reduce the traffic congestion. Researchers recently leveraged Deep Learning (DL) and Reinforcement Learning (RL) techniques to optimize TSCs. In RL framework, the agent interacts with surrounding world through states, rewards and actions. The formulation of these key elements is crucial as they impact the way the RL agent behaves and optimizes its policy. However, most of existing frameworks rely on hand-crafted state and reward designs, restricting the RL agent from acting optimally. In this paper, we propose a novel approach to better formulate state and reward definitions in order to boost the performance of the traffic signal controller agent. The intuitive idea is to define both state and reward in a consistent and straightforward manner. We advocate that such a design approach helps achieving training stability and hence provides a rapid convergence to derive best policies. We consider the double deep Q-Network (DDQN) along with prioritized experience replay (PER) for the agent architecture. To evaluate the performance of our approach, we conduct series of simulations using the Simulation of Urban MObility (SUMO) environment. The statistical analysis of our results show that the performance of our proposal outperforms the state-of-the-art state and reward design approaches.
Fetched live from OpenAlex and de-inverted. Abstracts are not stored in this database: the inverted indexes are 8.6 GB of the frame’s 9.3 GB of text, and the host has 13 GB free.
Full frame distilled prediction
Teacher imitationNot calibrated prevalence, not ground truth. Human validation pending. Learned from the 10,348 direct Codex labels and 10,348 direct Gemma labels. Candidate is the union of thresholded teacher heads; consensus is their intersection. These outputs are machine_predicted_unvalidated and are not human labels or direct frontier model labels.
Codex and Gemma teacher scores by category
| Category | Codex | Gemma |
|---|---|---|
| Metaresearch | 0.001 | 0.000 |
| Meta-epidemiology (narrow) | 0.000 | 0.000 |
| Meta-epidemiology (broad) | 0.000 | 0.000 |
| Bibliometrics | 0.000 | 0.000 |
| Science and technology studies | 0.000 | 0.000 |
| Scholarly communication | 0.000 | 0.000 |
| Open science | 0.000 | 0.000 |
| Research integrity | 0.000 | 0.000 |
| Insufficient payload (model declined to judge) | 0.000 | 0.000 |
Machine scores (provisional)
The two teacher heads of the student model, read on this work. A score orders the frame for review; it never asserts a category, and the validation status ships verbatim with every row.
Baseline scores from an immature model (maturity gate not passed, 7 training rounds). Scores rank; they never assert a category.
score_only:v0-immature-baseline · verbatim from the scoring run: score_only means the number may rank works, and no category label ships from it