Google: Autonomous Fleet

Fleet Management Dashboard for Autonomous Cabs
Designing a real-time fleet operations dashboard for autonomous vehicles, integrating live tracking, assignment logic, and system health monitoring. A 6-month capstone project embedded with Google.
The Problem: Disconnected Tools, Delayed Decisions
As autonomous vehicle fleets move from pilot programs to scalable reality, one question remains: how will operators manage them day to day? Unlike human-driven fleets, autonomous systems generate constant streams of data, but existing tools aren't built to help teams make sense of it in real time.
At 6 p.m. in a rush-hour simulation, 200 autonomous cabs crawl through downtown SF. Dispatchers juggle ride requests, blocked roads, and battery warnings, using three disjointed tools and half a dozen browser tabs.
The Solution: One Dashboard, Zero Guesswork
We designed a simulated fleet management system that brings everything together in a single command center. It empowers operators to respond faster, reroute smarter, and maintain vehicles more efficiently.
Key Features
Tracking Maintenance Without Spreadsheets
A maintenance board that tracks issue status (Open → In Progress → Resolved), paired with real-time health indicators and fleet stats. Drag-and-drop ticket updates, average resolution time of 4.2 hours, and 97.3% fleet uptime.
Making Alerts Instantly Actionable
A real-time feed of critical issues — battery, maintenance, routing, or connectivity — linked to quick actions. Replaced scattered warnings with one centralized, color-coded panel. Each alert opens a vehicle modal with status and instant options. 1-click triage, 22% faster fault resolution, no more missed alerts.
Smarter Route Optimization
A module that compares current vs. optimal route times using real-time + historical traffic data. Saved an average of 5.7 minutes per ride. One-click "Apply Best Route" to push to live dispatch.
User Research
We spoke to dispatchers, fleet supervisors, and maintenance leads. In total: 10 stakeholder interviews, 3 usability walkthroughs, and 1 card sort focused on alert prioritization.
"I'm jumping between five tools just to answer a simple question: is this car okay?" — fleet operator
Operators lacked a single source of truth. Dispatchers needed visibility to build trust. Maintenance workflows were fragmented. Map-based interactions were a common request.
Design Process
We started with low-fidelity wireframes to test layout logic and hierarchy. Each screen was designed to answer a specific question: Fleet Overview (Is my fleet running smoothly?), Alert Panel + Vehicle Modal (What needs attention?), and Maintenance Dashboard (Where are we in the resolution process?).
Final Design
With the layout locked, we shifted focus to how elements behave in a live environment. Design principles: fast cognition (alerts pulse, KPIs update without reloads), seamless flow (clicking an alert leads directly to a vehicle modal), and minimal friction (most actions are just one or two clicks).
Reflection
This project was a simulation, but it surfaced real design principles: design for action under pressure, not just visibility. Validate clarity through task-based testing, not just screens. Center workflows around the humans behind the dashboards. If I had more time, I would explore scaling to 10× more vehicles, add an AI co-pilot mode for proactive issue surfacing, and build a training flow for onboarding new dispatchers.