Skip to content
BranchesTags

Repository files navigation

JUNCTION

AI-Powered Automatic Block Planning to Maximize Asset Availability for Train Operations on Indian Railways

Smart India Hackathon 2026 Problem Statement ID: SIH26027 Ministry of Railways Live Prototype Tech Stack


"One system that sees the whole railway, not just one department's slice of it."

Live Prototype • Solution Overview • System Architecture • Technical Stack • 11-Step Workflow • Team

Executive Summary

Attribute Details
Problem Statement ID SIH26027
Problem Statement Title AI-Powered Automatic Block Planning to Maximize Asset Availability for Train Operations on Indian Railways
Theme & Category Transportation & Logistics | Software
Organization Ministry of Railways, Government of India
Live Prototype URL https://sihjunction.vercel.app
Official Repository https://github.com/samyakmisal/junction.git

The Problem & Our Solution

The Problem in Indian Railways Today

Fixed-infrastructure maintenance across Indian Railways is currently planned in departmental silos:

  • Civil Engineering (Track) logs track geometry, switch, and rail defects in TMS (Track Management System).
  • Electrical (TRD / OHE) logs 25kV catenary wear and transformer health in TDMS (Traction Distribution Management System).
  • Signalling & Telecom (S&T) manages point machines and axle counters in SMMS (Signalling Maintenance & Management System).
  • Train Operations manages dynamic train paths and section capacity in COA (Control Office Application).
  [TMS (Track)]        [TDMS (OHE)]        [SMMS (S&T)]        [COA (Timetables)]
         │                   │                   │                    │
         └───► Fragmented Block Requests (BDMS / Phone / Paper) ◄─────┘
                             │
                             ▼
  [!] Repeated Line Closures (Track blocked for Track work today, OHE work tomorrow)
  [!] Severe Passenger Train Delays (Vande Bharat / Rajdhani detained behind maintenance)
  [!] Suboptimal Crew & Heavy Machinery (Duomatic / BCM) Utilization

The Junction Solution

JUNCTION unifies maintenance data, sensor condition streams, and train timetables into a single intelligent planning and decision-support platform. It groups compatible tasks from multiple departments into coordinated "Shadow Blocks", schedules them during natural traffic lulls using Constraint-Based AI Optimization, and guarantees transparent, explainable recommendations with human-in-the-loop approval.

       TMS (Track)  +  TDMS (OHE)  +  SMMS (S&T)  +  COA (Timetables)
                                  │
                                  ▼
           ┌──────────────────────────────────────────────┐
           │                  JUNCTION                    │
           │  Unified Ingestion • MILP Optimizer • XAI    │
           └──────────────────────────────────────────────┘
                                  │
                                  ▼
  [+] Coordinated "Shadow Blocks" (3 department tasks collapsed into 1 window)
  [+] Zero Passenger Train Punctuality Loss (Targeted night & traffic lull slots)
  [+] Transparent AI (SHAP-style explainable reasoning for Chief Controllers)

How JUNCTION Addresses Core Challenges

┌──────────────────────────────────────┐     ┌────────────────────────────────────────────────────────┐
│ Challenge Anticipated                │ ──► │ How JUNCTION Responds                                  │
├──────────────────────────────────────┤     ├────────────────────────────────────────────────────────┤
│ Fragmented Planning Across 4 Systems │ ──► │ Unified Data Layer standardizes TMS, SMMS, TDMS & COA  │
│ Competing Departmental Priorities    │ ──► │ AI Priority Engine ranks urgency, risk & overdue score │
│ Repeated Track Closures              │ ──► │ Multi-Dept Coordination groups jobs into shared blocks │
│ Train-Maintenance Corridor Clashes   │ ──► │ Constraint Engine schedules windows with zero delays   │
│ Last-Minute Delays & Emergencies     │ ──► │ Real-Time Dynamic Re-planning & SLW Sandbox            │
└──────────────────────────────────────┘     └────────────────────────────────────────────────────────┘

Key Innovations & Uniqueness

  1. Predictive Maintenance Intelligence:
    • Analyzes cumulative Gross Million Tonnes (GMT), Track Geometry Index (TGI), Ultrasonic Flaw Detection (USFD), and Oscillation Monitoring System (OMS) readings to trigger maintenance before physical breakdown occurs.
  2. Multi-Department "Shadow Block" Clustering:
    • Automatically detects when Engineering, OHE, and S&T need access to the same physical section (e.g., KM 127/0 to 128/0) and consolidates them into a single track possession.
  3. Mathematical Constraint Optimization:
    • Employs Google OR-Tools (CP-SAT / MILP) to solve multi-objective trade-offs between train delay minimization, asset failure risk, and machine depot logistics.
  4. Explainable Railway Intelligence (XAI):
    • Replaces black-box AI with transparent reasoning. Displays SHAP-style feature importance and impact breakdowns so Controllers understand why a slot was chosen.
  5. Human-in-the-Loop Governance:
    • No schedule is forced without human authority. Controllers can Approve, Modify Time Windows, or Reject with one click.
  6. Real-Time What-If Emergency Sandbox:
    • Simulates sudden incidents (Rail Fracture, OHE Catenary Snap, Freight Surges) and immediately outputs Single Line Working (SLW) and Temporary Speed Restriction (TSR) diversion directives.

System Architecture

graph TD
    subgraph Data_Sources ["1. Data Ingestion & Normalization Layer"]
        TMS["TMS<br/>(Civil Track)"]
        TDMS["TDMS<br/>(OHE & SCADA)"]
        SMMS["SMMS<br/>(Signalling & Telecom)"]
        COA["COA & FOIS<br/>(Train Timetables)"]
        Sensors["Sensors / USFD / TRC<br/>(Asset Telemetry)"]
    end

    subgraph Core_Engine ["2. Unified Intelligence & Optimization Engine"]
        Ingest["Unified Normalizer & PostGIS Spatial Layer"]
        Priority["Asset Risk & Urgency Engine (XGBoost / Rules)"]
        Optimizer["MILP Constraint Optimizer (Google OR-Tools CP-SAT)"]
        Conflict["Conflict Detection & Shadow Clustering Engine"]
    end

    subgraph Decision_Support ["3. Visualization & Decision Interface"]
        Gantt["Interactive Gantt Timeline (00:00 - 08:00)"]
        Map["Spatial Live Network Map (UP/DN/Loops)"]
        XAI["Explainable AI Decision Synthesizer"]
        Sandbox["What-If Emergency Simulation Sandbox"]
    end

    subgraph Human_Loop ["4. Governance & Execution"]
        Controller["Chief Controller Review (Approve / Modify / Reject)"]
        Exec["Approved Block Notice & Speed Restriction (TSR)"]
        Audit["Tamper-Evident Regulatory Audit Trail"]
    end

    Data_Sources --> Ingest
    Ingest --> Priority
    Priority --> Optimizer
    COA --> Optimizer
    Optimizer --> Conflict
    Conflict --> Gantt & Map & XAI & Sandbox
    Gantt & XAI --> Controller
    Controller -->|Approve| Exec
    Controller -->|Override| Optimizer
    Exec --> Audit
Loading

End-to-End Conceptual Demo Flow

The 11-step lifecycle of how JUNCTION turns raw condition signals into an executed block:

 ┌─────────────┐     ┌─────────────┐     ┌─────────────┐     ┌─────────────┐     ┌─────────────┐
 │ 1. ASSET    │ ──► │ 2. MAINT.   │ ──► │ 3. RAILWAY  │ ──► │ 4. DEPT.    │ ──► │ 5. RULE &   │
 │    ALERT    │     │    DECISION │     │    CHECK    │     │    CHECK    │     │    CONFLICT │
 └─────────────┘     └─────────────┘     └─────────────┘     └─────────────┘     └─────────────┘
  Condition drops     Is block needed?    Check train         Civil, OHE & S&T    Safety, track &
  GMT / TGI alert     Duration estimate   timetable & slots   find overlaps       isolation rules
                                                                                        │
                                                                                        ▼
 ┌─────────────┐     ┌─────────────┐     ┌─────────────┐     ┌─────────────┐     ┌─────────────┐
 │ 10. FINAL   │ ◄── │ 9. HUMAN    │ ◄── │ 8. RECOMMEN-│ ◄── │ 7. TRAIN    │ ◄── │ 6. FIND BEST│
 │     PLAN    │     │    DECISION │     │    DATION   │     │    IMPACT   │     │    WINDOW   │
 └─────────────┘     └─────────────┘     └─────────────┘     └─────────────┘     └─────────────┘
  Time, Section &     Approve / Modify    XAI reasoning &     Delays & diversion  OR-Tools CP-SAT
  Possession Memo     or Reject by Chief  confidence score    calculation         slot evaluation
        │
        ▼
 ┌─────────────┐
 │ 11. EXECUTE │ ──► [Continuous Feedback Loop: Updates AI models with actual execution data]
 └─────────────┘

Application Features & Module Walkthrough

1. Operations HUD Dashboard

  • Corridor Health Metric: Real-time aggregation of track reliability (e.g. 94.2%).
  • Active KPI Counters: Active blocks, speed restrictions (TSRs), impending conflicts, and train throughput.
  • Emergency Quick-Trigger: 1-click corridor-wide emergency simulation for rapid disaster drill response.

2. AI Multi-Horizon Block Optimization Engine

  • Multi-Horizon Support:
    • 24H Dynamic Horizon: Real-time dispatching and immediate conflict avoidance.
    • 7-Day Tactical Horizon: Weekly machine gang (Duomatic / BCM) roster planning.
    • 30-Day Strategic Horizon: Major track renewal, turnout replacement, and bridge rehabilitation.
  • Tuning Objective Weights:
    • $w_1$: Train Delay Minimization
    • $w_2$: Asset Failure Risk Urgency
    • $w_3$: Multi-Department Shadow Clubbing Priority
    • $w_4$: Machine & Crew Depot Availability

3. Interactive Gantt Schedule Visualizer

  • Time Scale Matrix (00:00 to 08:00): Visualizes train movements (Vande Bharat, Rajdhani, BOXN Goods) against departmental block proposals.
  • Live "NOW" Marker: Real-time tracking of corridor time.
  • Optimal AI Slot Highlight: Visually displays the zero-delay maintenance window.

4. Asset Intelligence & Explainable AI (XAI)

  • Kilometer-Wise Granularity: Exact asset pins (e.g., Turnout 101 at KM 127/4 on UP Line).
  • Engineering Metrics: 60kg/52kg Rail classification, GMT load, TGI, USFD flaw status, OHE stagger, and Point stroke times.
  • SHAP-Style Feature Impact: Clear ranking of top contributing failure risk factors.

5. Conflict Resolution Center

  • Automatically flags:
    • Overlapping track possessions across departments.
    • Train corridor timetable clashes.
    • 25kV OHE power isolation boundary conflicts.
    • Peak-hour commuter schedule violations.

6. What-If Emergency Simulation Sandbox

  • Simulates live crises:
    • Rail Fracture (KM 127/4 UP Line)
    • 25kV OHE Catenary Dropper Parting
    • Goods Freight Surge (+40% traffic)
  • Automatically generates Single Line Working (SLW) protocols, holding patterns, and temporary speed restrictions with zero passenger disruption.

7. Dedicated Department Workspaces

  • Track Department: Manual inspection logging, sleeper and ballast status, USFD crack alerts.
  • OHE Department: Contact wire wear %, SCADA isolator state, dropper tension.
  • S&T Department: Point machine operating time, track circuit drop status, axle counter health.

8. Regulatory Compliance & Audit Log

  • Records every controller sanction, time override, emergency trip, and digital token generation with timestamps and role identifiers.

Technical Stack

┌────────────────────────────────────────────────────────────────────────┐
│                              FRONTEND                                  │
├────────────────────────────────────────────────────────────────────────┤
│ • React 18 (TypeScript)         • Tailwind CSS (Tactical Dark HUD)     │
│ • Vite (Ultra-fast bundler)     • Lucide React (Industrial Icons)      │
│ • Recharts (Condition Charts)   • Canvas-Confetti (Interactive UX)     │
├────────────────────────────────────────────────────────────────────────┤
│                              BACKEND                                   │
├────────────────────────────────────────────────────────────────────────┤
│ • Python FastAPI (Async API)    • Supabase Auth (RBAC & Profiles)      │
│ • WebSockets (Live Telemetry)   • Pydantic (Type-safe Schemas)         │
├────────────────────────────────────────────────────────────────────────┤
│                     OPTIMIZATION & AI PIPELINE                         │
├────────────────────────────────────────────────────────────────────────┤
│ • Google OR-Tools (CP-SAT)      • Mixed-Integer Linear Prog. (MILP)    │
│ • XGBoost / Scikit-Learn        • Explainable AI (SHAP-style reasoning)│
├────────────────────────────────────────────────────────────────────────┤
│                        DATABASE & SPATIAL GIS                          │
├────────────────────────────────────────────────────────────────────────┤
│ • Supabase (PostgreSQL with RLS)• PostGIS (Kilometer & Track Chainage) │
│ • TimescaleDB (Time-series)     • Redis (Sub-second Conflict Cache)    │
└────────────────────────────────────────────────────────────────────────┘

Impact & Measurable Benefits

┌────────────────────────────┐  ┌────────────────────────────┐  ┌────────────────────────────┐
│   OPERATIONAL & SOCIAL     │  │         ECONOMIC           │  │       ENVIRONMENTAL        │
├────────────────────────────┤  ├────────────────────────────┤  ├────────────────────────────┤
│ • Punctuality improvement  │  │ • Avoidable downtime cut   │  │ • Reduced diesel loco      │
│   for Superfast & Express  │  │   by up to 28%             │  │   relocations              │
│ • Lower emergency failures │  │ • Heavy machine (BCM /     │  │ • Decreased fuel burn      │
│ • Unified inter-dept       │  │   Duomatic) ROI to 92%+    │  │   from freight idling      │
│   possession planning      │  │ • Reduced detention costs  │  │ • Sustainable operations   │
└────────────────────────────┘  └────────────────────────────┘  └────────────────────────────┘

Key Impact Highlights:

  • Fewer Repeated Blocks: Bundling compatible jobs cuts redundant possessions by over 35%.
  • Higher Asset Availability: Maximize track uptime across high-density corridors.
  • Lower Train Disruption: Dynamic re-routing preserves passenger train timetables.
  • Fail-Safe Decision Support: Transparent AI explanations with mandatory human controller sign-off.

Field Research & Railway References

Our architecture and domain models are built directly on real-world Indian Railways practices:

  • Primary Field Research: Conducted railway field visits and in-depth interviews with Track Engineers, OHE Traction Staff, and Section Controllers.
  • Official Reference Manuals:
    • Indian Railways Permanent Way Manual (IRPWM)
    • Manual of Instructions on Track Recording Car (TRC)
    • OHE, PSI & SCADA Maintenance Manual for Traction Distribution (TRD)
    • Indian Railways Signal Engineering Manual (IRSEM)
  • Mathematical Foundations: Google OR-Tools Constraint Programming (CP-SAT) formulation for Job-Shop Scheduling with Spatial Track-Occupancy Constraints.

Getting Started (Local Development)

Prerequisites

  • Node.js: v18.0.0 or higher
  • npm: v9.0.0 or higher (or pnpm / yarn)

Quick Setup

  1. Clone the Repository:

    git clone https://github.com/samyakmisal/junction.git
cd junction

  2. Install Dependencies:

    npm install

  3. Start Local Development Server:

    npm run dev

    Open http://localhost:3000 in your browser.

  4. Build for Production:

    npm run build

  5. Preview Production Build:

    npm run preview

Team Junction

Samyak Misal
Team Lead		 Vishv Chavan
Member 1		 Gauri Gandre
Member 2
Full-Stack Architecture & Optimization Engine Frontend UI/UX & Spatial Network GIS Asset Deterioration & Predictive ML
Sai Dhapte
Member 3		 Suraj Kolpe
Member 4		 Sourabh Patil
Member 5
Railway Domain Research & Track/OHE Modeling FastAPI Backend & OR-Tools Solver Integration Conflict Matrix, QA & Field Validation

Smart India Hackathon 2026 • Ministry of Railways (PS ID: SIH26027)

About

AI-Powered Automatic Block Planning & Multi-Department Shadow Scheduling System for Indian Railways (SIH 2026 • PS: SIH26027).

junction-taupe.vercel.app

Topics

block-planningexplainable-aifastapigoogle-or-toolsindian-railwayspostgispredictive-maintenancerailway-operationsreactsih2026smart-india-hackathontypescript

Resources

Readme
Activity

Stars

2 stars

Watchers

0 watching

Forks

0 forks
Report repository

Releases

Packages

Contributors

Languages