Stadium Flow Advisory — Real-Time Crowd Optimization System

Problem Context

Large-scale stadium events consistently face uneven crowd distribution at entry points. Certain gates become heavily congested while others remain underutilized, leading to increased wait times, potential safety risks, and inefficient use of infrastructure. The core issue is not lack of capacity, but lack of real-time, personalized routing intelligence.

---

System Overview

Stadium Flow Advisory is a real-time decision system that dynamically routes users to the most efficient entry gate. Instead of relying on static proximity (nearest gate), the system continuously computes an optimal entry path based on both spatial and temporal factors.

The platform operates as a feedback-driven system where:

• User location provides spatial input
• Crowd reports and organizer inputs provide temporal congestion data
• The backend computes optimal routing decisions
• Results are pushed to users in real time

---

Core Algorithm: Optimal Gate Selection

Each gate is modeled as a queueing system. The objective is to minimize total entry time rather than distance.

For a given user location:

• Walking Distance is calculated using OSRM based on real road/footpath networks
• Walking Time is derived from distance using an average walking speed (~1.4 m/s)
• Queue Time is estimated from live crowd density signals

The system computes:

Total Cost = Walking Time + Queue Delay

This converts the problem into a multi-factor optimization task rather than a simple nearest-neighbor lookup.

The system evaluates all candidate gates and selects the one with the minimum total cost. This decision is recalculated continuously as new data arrives.

---

Real-Time Crowd State Estimation

Crowd density is not directly measured but inferred using a hybrid model:

1. Crowd-Sourced Reports
   • Users submit local observations (e.g., sparse, moderate, dense)
   • Reports are geo-tagged and time-stamped
2. Validation Layer
   • Nearby users can verify reports
   • Confidence scores are assigned based on agreement
3. Aggregation Layer
   • Reports are cached and aggregated (Redis-ready design)
   • Gate-level crowd states are derived from weighted averages
4. Organizer Overrides
   • Organizers can directly set or adjust crowd levels and queue estimates

This creates a self-correcting system where inaccurate data is naturally filtered out over time.

---

Routing Engine Integration

The system integrates with OSRM to perform:

• Shortest path computation for pedestrian routes
• Distance and duration estimation
• Route geometry generation for frontend visualization

This ensures routing decisions are based on actual walkable paths, not straight-line approximations.

---

Incentive Mechanism for Load Balancing

A key challenge is behavioral: users prefer the nearest gate even if it is slower overall.

To address this, the system introduces a detour incentive model:

• If a user is recommended a farther but faster gate
• And the user chooses to follow that recommendation
• The system tracks completion (arrival at gate region)
• Rewards are issued via backend validation

This transforms routing from passive suggestion into active crowd redistribution.

---

Dual Interface Architecture

Fan Interface

• Real-time map visualization (Leaflet + OpenStreetMap)
• Gate markers with dynamic attributes:
  • Queue time
  • Walking time
  • Total estimated entry time
• Route rendering for selected gate
• Lightweight reporting interface for crowd input

Organizer Interface

• Centralized control panel
• Gate state management (open/close/emergency)
• Real-time crowd distribution overview
• Broadcast system for instant updates
• Report verification tools

---

Privacy Model

The system is designed with minimal data retention:

• GPS data is processed in real time and not permanently stored with user identity
• Crowd insights are aggregated at gate-level granularity
• No personally identifiable tracking is required for system functionality

This ensures operational efficiency without compromising user privacy.

---

System Architecture

Frontend

• Next.js with React and TypeScript
• Leaflet for map rendering
• State-driven UI for real-time updates

Backend

• Node.js with Express
• REST APIs for core operations
• Designed to support WebSocket integration for live updates

Data Layer

• SQLite (sql.js) for lightweight persistence
• Suitable for hackathon deployment and edge scenarios

Routing Layer

• OSRM for geospatial computation

---

API Design

• GET /api/gates/recommendation
  Returns optimal gate based on user coordinates
• GET /api/reports/live
  Returns aggregated crowd state per gate
• POST /api/reports
  Submit crowd observation
• POST /api/rewards/detour-points
  Validate and assign rewards
• POST /api/updates
  Organizer broadcast endpoint

---

Key Outcomes

• Reduces average entry time by distributing crowd load
• Prevents localized congestion and potential safety risks
• Provides organizers with real-time operational control
• Encourages cooperative user behavior through incentives

---

Closing Thought

This system shifts crowd management from reactive monitoring to predictive and adaptive control, using real-time data and optimization logic to continuously improve flow across the stadium.