Research report · Active
SafeCrossAI: Intelligent Intersection Safety
AI mobility research · Active · 2025–Present
SafeCrossAI investigates how infrastructure-based AI can understand multi-agent traffic scenes, predict vulnerable road-user motion, and estimate safety-critical interactions at smart intersections.
Abstract
Develop a research platform for vulnerable road-user trajectory prediction, interaction modelling, risk assessment, and intelligent intersection safety.
Scientific question: Can infrastructure-based AI predict vulnerable road-user motion and expose uncertainty for safer intersection decisions?
Contribution: A smart-intersection research platform with baseline prediction, social features, and risk utilities.
Marked research artifact slot
Smart-intersection AI architecture
Architecture figure slot for verified diagrams of modules, data flow, and software boundaries.
Marked research artifact slot
Detect · predict · reason about risk
Pipeline slot for GIFs or animations that explain estimation, planning, perception, or evaluation steps.
Marked research artifact slot
Synthetic intersection safety scene
Demo slot reserved for real project videos or synthetic demos that are explicitly labeled as synthetic.
Research questions
- How can infrastructure perception improve vulnerable road-user trajectory prediction?
- Which social interaction features are useful for safety-critical forecasting?
- How can uncertainty-aware prediction support connected and autonomous mobility?
Methods
Implementation status
Implemented
- Constant-velocity baseline
- ADE/FDE evaluation
- Neighbor search
- Pairwise social-feature utilities
Research prototype
- Interaction graph construction
- Risk reasoning utilities
- Dashboard concept
Planned
- Public dataset loaders
- Graph neural predictors
- Uncertainty calibration report
Experiments and metrics
Experiments
- Constant-velocity baseline
- Neighbor search
- Pairwise social features
- Interaction graph construction
Metrics
- ADE
- FDE
- time-to-collision
- calibration error — planned
Quantitative benchmark tables will be added only after reproducible experiments are available.
Limitations
- Quantitative claims are intentionally withheld until reproducible benchmark runs are available.
- Visual figures and GIFs are placeholders unless a project page explicitly states that an artifact is generated from real experiments.
Reproducibility plan
Experiment workflow
Each module is prepared to document configuration files, datasets, evaluation metrics, and repeated experiment runs before quantitative claims are shown.
Repository setup
Public repositories are linked when available. Private, planned, or incomplete repositories remain clearly labelled rather than being presented as finished systems.
Literature context
- Trajectory forecasting
- Social interaction modelling
- VRU safety
- Intelligent transportation systems
Roadmap
- Add public dataset loaders
- Integrate graph neural predictors
- Build visualization dashboard