ParkWise
An interactive, browser-based 3D parking simulator that helps learner drivers practise the common exam maneuvers and build real spatial confidence — with physics-based car movement and a coach that guides each step.
ParkWise runs instantly in any modern browser. No install, no plugins, no account.
Inspiration
Last week my younger cousin told me how anxious she was about her upcoming driving road exam. Like a lot of new drivers, her biggest stress was parking: how to approach a space, when to turn the wheel, and how not to clip the curb.
The driver handbooks we looked at used flat 2D overhead diagrams that never really explain how a real car pivots around its rear wheels, and video tutorials gave her nothing to actually practise on. She's non-technical, so installing a game engine or heavy software was out. I wanted something she could open from a link and start using straight away — a safe visual sandbox to build spatial awareness before she's in a real car.
What it does
ParkWise is an interactive 3D spatial trainer for learner drivers:
- The common exam maneuvers — guided practice for parallel, reverse bay, and angle parking, with more situations I can add from my own driving experience as time allows.
- Realistic steering, not arcade rotation — the car moves on a kinematic bicycle model, so it pivots around the rear axle like a real car rather than spinning on its center.
- A coach that demonstrates, then guides — the coach first drives the maneuver itself so you can watch it, then you take the wheel and it tells you the distance to the car ahead/behind and how much to adjust, with reference lines on the road.
- Multiple camera views — overhead, chase, and a driver's-seat view with working mirrors, so you can judge your alignment from any angle.
- Proximity cues — parking-sensor-style beeps that speed up as you near a curb or another car.
- A "what not to do" demo — shows a common mistake and why it goes wrong, so you learn the failure as well as the success.
How I built it
I built ParkWise with the Devpost Learn plan-first, flipped-interaction workflow — scoping the idea, writing the requirements and a technical spec, then building it in verifiable steps, before and during the code.
- Planning first: the agent interviewed me and I made the decisions; each planning doc guided the next.
- 3D in the browser: the scene, lighting, and car are built with Three.js, with everything bundled locally so it works fully offline.
- Car physics: the steering and turning geometry are a hand-written kinematic bicycle model in vanilla JavaScript — no physics engine — which is light and accurate for low-speed parking.
- Audio: proximity cues are generated with the native Web Audio API, so there are no audio files to download.
- UI: a clean HUD in plain HTML and CSS, with locally bundled fonts.
Challenges
- Getting reverse steering right. Reversing is the counterintuitive part of the bicycle model and most of what parallel parking is, so I verified that behavior early before building the coach and scenarios on top of it.
- Keeping it smooth in the browser. Real-time 3D with shadows had to stay responsive on an ordinary laptop, which meant keeping the scene and draw calls in check.
- A readable HUD. Showing useful steering cues and distances without covering the parking view took some back-and-forth on the camera and overlay.
What I learned
- How vehicle turning mechanics translate into interactive 3D code — writing movement grounded in real physics.
- How to work with an AI coding agent through structured planning instead of unstructured prompting, and the value of keeping some decisions deliberately open until build time.
- Practical techniques for 3D performance in the browser.
What's next
- Different vehicle sizes, to show how wheelbase affects the turning point.
- More scenarios drawn from my own driving experience.
- Printable one-page maneuver summaries for quick review before test day.
Run it locally
Requires Node.js (v18+).
cd parkwise
npm start
Then open http://localhost:3000.
Planning documents
Built following the Devpost Learn plan-first workflow. The planning artifacts live in devpost/:
Tech stack
- 3D: Three.js (WebGL)
- Physics: hand-written kinematic bicycle model
- Audio: Web Audio API (procedural)
- UI: vanilla HTML and CSS, locally bundled fonts
- Server: zero-dependency Node static server
License
MIT