Research map / Reinforcement learning / Offline RL
EcoLight: Intersection Control in Developing Regions Under Extreme Budget and Network Constraints
Sachin Chauhan, Kashish Bansal, Rijurekha Sen
Read the paperCheck your idea against 137,217 papers
AI-extracted. The measured effects below were read by AI from the paper's full text (no summary yet). Numbers and claims can be wrong or missing context, so check the paper before you cite or build on it.
Measured effects reported in the paper
| What was changed | Where | Measured | Effect |
|---|---|---|---|
| use 1-dimensional state instead of 2-dimensional state in DRL (ablation) | 2dimRL · 16x3 NY | nOut: 2574 -> 1148 | Hurt |
| use 1-dimensional state instead of 2-dimensional state in DRL (ablation) | 2dimRL · 16x3 NY | Total: 322 -> 1093 | Hurt |
| use 1-dimensional state instead of 2-dimensional state in DRL (ablation) | 2dimRL · 1x1 Delhi | nOut: 1282 -> 1260 | Hurt |
| use 2-dim state DRL (162 params) instead of CoLight (6018 params) | CoLight · 16x3 NY | nOut: 2589 -> 2574 | No clear effect |
| quantize traffic density into a lookup table with goodness selection (DecisionConsistency TS) vs 2dimRL | 2dimRL · 1x1 Delhi | nOut: 1282 -> 1298 | Improved |
| Decision(TS) vs 2dimRL on travel time | 2dimRL · 1x1 Delhi | Travel: 237.9 -> 186.4 | Improved |
| add goodness-based selection (TS) to training-only (T) for DecisionConsistency (ablation) | Decision · 16x1 NY | nOut: 5081 -> 5186 | Improved |
| add goodness-based selection (TS) to training-only (T) for DecisionConsistency (ablation) | Decision · 16x3 NY fairness | WrstTime: 924.7 -> 675.2 | Improved |
| add goodness-based selection (TS) to training-only (T) for FairShare (ablation) | FairShare · 16x3 NY fairness | WrstTime: 1207.1 -> 929.0 | Improved |
| use DecisionConsistency instead of FairShare goodness metric (TS) (ablation) | LUT EcoLight · 1x1 Delhi | nOut: 1285 -> 1298 | Improved |
| replace CoLight with Decision(TS) LUT on fairness | CoLight · 16x3 NY fairness | WrstTime: 834.4 -> 675.2 | Improved |
| stateless Timed(2dim) vs SOTL baseline | SOTL · 16x3 NY | nOut: 2462 -> 2553 | Improved |
Closest papers by meaning
- IntersectionZoo: Eco-driving for Benchmarking Multi-Agent Contextual Reinforcement Learning (ICLR 2025 · similarity 0.87)
- RobustLight: Improving Robustness via Diffusion Reinforcement Learning for Traffic Signal Control (ICML 2025 · similarity 0.85)
- π-Light: Programmatic Interpretable Reinforcement Learning for Resource-Limited Traffic Signal Control (AAAI 2024 · similarity 0.84)
- DiffLight: A Partial Rewards Conditioned Diffusion Model for Traffic Signal Control with Missing Data (NeurIPS 2024 · similarity 0.84)
- VLMLight: Safety-Critical Traffic Signal Control via Vision-Language Meta-Control and Dual-Branch Reasoning Architecture (NeurIPS 2025 · similarity 0.84)
- CoLLMLight: Cooperative Large Language Model Agents for Network-Wide Traffic Signal Control (ICLR 2026 · similarity 0.82)
Writing a paper in this area? scime draft is a free LaTeX editor with AI included, real-time co-authors and a check of your references and venue rules before you submit. It also imports OpenAI Prism projects.
