Optimizing Pedestrian Safety in Real-Time: An Extreme Value Theory-Based Reinforcement Learning Framework
Bibliographic record
Abstract
As cities continue to encourage active transportation, increased focus has been placed on pedestrian safety given their vulnerability and proneness to more severe crashes in environments where they coexist with other road users. Many cities have introduced pedestrian-friendly designs, such as raised crosswalks, narrowed lanes, and signal timing changes like Leading Pedestrian Intervals (LPIs), which allow pedestrians to enter intersections ahead of vehicles. This improves visibility, minimizes the number of unsafe interactions, and reduces the crash risk overall. Simultaneously, cities have been preparing for the massive introduction of connected and autonomous vehicles (CAVs), which will use vehicle-to-everything (V2X) communications to interact with infrastructure and enable real-time traffic optimization through Actuated Traffic Signal Controls (ATSCs). Despite these advances, few studies have focused on incorporating pedestrian safety metrics into the operation and optimization of advanced ATSCs. This work develops a novel framework that integrates real-time crash risk metrics using Extreme Value Theory (EVT) with a Reinforcement Learning (RL) controller that dynamically adjusts signal timings based on real-time traffic conditions. The proposed system selects between introducing an LPI or maintaining the standard phase sequence at each signal cycle by using a multi-objective function that considers both safety and mobility. Case studies at two intersections in Vancouver, Canada, show that the approach can substantially improve pedestrian safety while preserving acceptable traffic performance levels for all users. This is especially useful for locations with high pedestrian activity, where the algorithm is able to improve their safety without compromising mobility.
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.
How this classification was reachedexpand
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 | 0.007 | 0.002 |
| Meta-epidemiology (narrow) | 0.001 | 0.001 |
| Meta-epidemiology (broad) | 0.001 | 0.000 |
| Bibliometrics | 0.001 | 0.004 |
| Science and technology studies | 0.001 | 0.000 |
| Scholarly communication | 0.002 | 0.002 |
| Open science | 0.003 | 0.001 |
| Research integrity | 0.001 | 0.003 |
| Insufficient payload (model declined to judge) | 0.002 | 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 itClassification
machine, unvalidatedMachine predicted; a candidate call from one teacher head, not a consensus.
How this classification was reached, model by model and score by score, is at the end of the page under "How this classification was reached".