PREDISS: Predictive Intersection Safety System
PREDISS (Predictive Intersection Safety System) is a state-of-the-art solution designed to enhance safety for vulnerable road users (VRU) at signalized intersections. Developed by a collaborative team from FAMU-FSU College of Engineering and Tallahassee Advanced Traffic Management System, PREDISS leverages machine learning, controls, optimization, and connected vehicle technologies to create a comprehensive intersection monitoring system.
Technical Overview
The system employs an advanced multi-stage fusion framework that processes data from multiple sensors including cameras, LiDAR, inductive loops, and radar. Key technical features include:
- Multi-sensor fusion architecture combining visual cameras, LiDAR, and thermal cameras
- Advanced object detection and classification
- Real-time trajectory prediction and conflict assessment
- Joint Probabilistic Data Association (JPDA) framework with Kalman filtering for robust tracking
- Class-specific fusion strategies utilizing spatial indexing
- Specialized prediction methodologies for different road user classes, with physics-based constraints
Implementation Features
PREDISS stands out for its practical implementation approach, designed for real-world deployment across existing intersections. Notable features include:
- Modular architecture adaptable to various sensor configurations and intersection layouts
- User-friendly GUI-based calibration system for quick setup and configuration
- Computationally efficient algorithms enabling real-time processing on modest hardware
- Adaptive fusion techniques compatible with existing municipal infrastructure
- Minimal setup requirements using readily available satellite imagery and city maps
News
January 2025 – PREDISS Wins DOT Intersection Safety Challenge
The FAMU-FSU team’s PREDISS system has been selected as a winner in the U.S. DOT Intersection Safety Challenge Stage 1B: System Assessment and Virtual Testing Primary Track. The announcement was made at the 2025 Transportation Research Board (TRB) Annual Meeting. The challenge focused on developing intersection safety systems leveraging AI and machine learning to identify and mitigate unsafe conditions involving vehicles and vulnerable road users. The team’s solution demonstrated exceptional performance in sensor fusion, classification, and conflict prediction using real-world sensor data collected at the FHWA Turner-Fairbank Highway Research Center.