Fleet Telematics & Driver Safety for a 1,800-Truck Logistics Fleet
Course: B.Tech CSE · Software Engineering & Project Management (SEPM)
Case Study: No. 19 - HaulTrack
Student: Vedh Naik (Roll No:150096725163) · Cohort: Jensen Huang
👉 Launch Interactive Web Demo (GitHub Pages)
HaulTrack is a fleet telematics and driver-safety system for a logistics firm with 1,800 trucks, each reporting location, speed, harsh braking, idling and fuel level every 30 seconds. One system has to satisfy three stakeholders with three definitions of success:
| Stakeholder | Wants | HaulTrack's answer |
|---|---|---|
| Fleet manager | Route deviation alerts; a driver safety score | Dispatcher alerts with suppression rules; an advisory coaching score |
| Drivers' union | Assurance the score cannot be used for arbitrary dismissal | A written Data Use Policy: the score is never the basis for discipline |
| Insurer | Evidence that harsh braking has fallen | A frozen, terrain-normalised harsh-braking KPI served as aggregates |
The core engineering lesson:
"The previous vendor's score penalised drivers on hill routes. That was a requirements failure, not an algorithm failure."
So fairness is written as measurable requirements (TN-1 to TN-7) and enforced by a regression test suite that fails the old algorithm.
| Item | Result |
|---|---|
| Intermediate COCOMO (44 KLOC, semi-detached, EAF 1.15) | 239.0 person-months, about 17.0 months, about 14 FTE average |
| Basic COCOMO (same size) | 207.9 person-months, about 16.2 months |
| Difference | +31.2 PM (+15%), all from the EAF |
| Daily ingestion | 1,800 × 2,880 = 5,184,000 records/day (about 60/s average) |
| Retention | Raw 90 days · identified events 13 months · pseudonymised aggregates 36 months |
Truck unit -> Secure gateway -> Queue -> Validate/dedupe -> Enrich (map-match + road grade)
-> Event detector (terrain-aware) -> Nightly scoring (28-day window)
-> Manager console / Driver self-view / Insurer aggregates API
-> Route deviation detector -> Dispatcher alerts (never feeds the score)
Terrain-aware harsh-braking thresholds (net deceleration, downhill bands): under 2% = 3.0 · 2 to under 5% = 3.3 · 5 to under 8% = 3.6 · 8% and above = 4.0 m/s².
| # | Deliverable | File | Key contents |
|---|---|---|---|
| 1 | Stakeholder Conflict Analysis | 01_Stakeholder_Conflict_Analysis_HaulTrack.pdf |
Stakeholder register, 6 conflicts, gate + weighted-scoring prioritisation, 6 resolution decisions, terrain requirements TN-1..TN-7 |
| 2 | Estimation Report | 02_Estimation_Report_HaulTrack.pdf |
Intermediate COCOMO, 15-driver table with justification, basic-model comparison |
| 3 | Design Pack | 03_Design_Pack_HaulTrack.pdf |
Ingestion sizing, pipeline architecture, scoring model, class diagram, route deviation activity diagram |
| 4 | Test Design | 04_Test_Design_HaulTrack.pdf |
Terrain scoring tests, route deviation tests, hill-route regression (HR-01..05), policy tests |
| 5 | Project Plan | 05_Project_Plan_HaulTrack.pdf |
WBS, Gantt (74 weeks), communication plan, union concern as a risk, risk register |
| 6 | Data Use Policy | 06_Data_Use_Policy_HaulTrack.pdf |
Retention, access control matrix, permitted and prohibited uses, appeals, governance |
| 7 | Interactive Telematics Demo | Live Web Demo · index.html |
Standalone interactive UI demo: Terrain-Aware vs Legacy Scorer, Route Deviation Simulator with canvas, Data Use Policy Engine, Fleet Sizing |
- 🌐 Live Demo (No installation required): https://voidvedh.github.io/HaulTrack/
- 💻 Local Access: Open
index.htmldirectly in any modern web browser.
The web app demonstrates all core SEPM requirements through interactive widgets:
- 🏔️ Terrain-Aware vs. Legacy Scorer: Interactive grade-band and deceleration sliders comparing the fair, physics-grounded HaulTrack model against the unfair legacy vendor scoring that penalized hill drivers. Includes presets for key test scenarios (ST-01 flat-road harsh brake, ST-04 6% grade mitigation, ST-05 steep 9% grade, ST-07 unknown gradient, ST-12 low-mileage dampening, ST-13 score floor).
- 🗺️ Route Deviation Canvas Simulator: Live HTML5 canvas tracking a delivery truck along a 500m safe corridor with 30-second tick steps, consecutive off-corridor notice transitions (0/3 to 3/3), escalation to dispatcher alerts, and 5 operational suppression toggles (Depot Geofence, Planned Stop, Road Closure, Dispatcher Reroute, GPS HDOP > 5.0).
- 🛡️ Data Use Policy Engine: Interactive role & purpose evaluator testing governance rules (Driver Self-View, Manager Coaching, Insurer Aggregates, Disciplinary Score Export Blocking, and HR Panel Barred Use) with a live tamper-evident audit log.
- 📊 Fleet Architecture & Retention Lifecycle: Live pipeline sizing metrics for 1,800 trucks (5.18M pings/day), COCOMO project estimates (44 KLOC, 239 PM), and 90-day / 13-month / 36-month data lifecycle stages.
- Student: Vedh Naik
- Roll No.: 150096725163
- Cohort: Jensen Huang
- Course: B.Tech Computer Science & Engineering · Software Engineering & Project Management
- Problem Statement: Case Study No. 19 - HaulTrack
- Standards / models applied: COCOMO 81 (Boehm), UML 2, MoSCoW prioritisation, IEEE 829-style test design