TheFrontierMile

Inspiration: TheFrontierMile was inspired by my personal experience using Tesla Full Self-Driving (FSD). During autonomous-driving sessions, I've encountered moments when I wished the vehicle would make a slightly different decision. Not necessarily because its planned maneuver was unsafe, but because it didn't reflect my preferences as a passenger. For example, the car may not yield as quickly for pedestrians and create an awkward situation for me as a driver. In situations like these, I want to communicate my preference without physically taking over or disengaging the autonomous-driving session.

That experience inspired TheFrontierMile: a passenger-assisted autonomous-driving concept that introduces a way for people to provide contextual input while the vehicle continues driving autonomously.

What it does: TheFrontierMile is an interactive passenger-assisted autonomous-driving demo named PAADD, a passenger interface that allows users to request subtle adjustments to the vehicle's driving behavior without ending the autonomous-driving session.

The project demonstrates three scenarios in which a vehicle's planned behavior may be technically permissible but could benefit from additional human context:

1. Campus pedestrian crossing: A vehicle approaches pedestrians without slowing as early as the passenger would prefer. Using PAADD, the passenger requests an earlier, smoother yield.
2. School-zone speed adjustment: A vehicle approaches a simulated school zone with a reduced speed limit during designated hours. The passenger requests an adjustment to the appropriate speed without disengaging autonomy.
3. Highway lane change: A vehicle plans to merge from the right lane into an available center lane while a semi-truck travels alongside it in the far-left lane. Although the center lane is open, the passenger feels uncomfortable merging directly beside the truck. PAADD allows the passenger to defer the maneuver until the semi passes. Each interaction is evaluated by a simulated driving supervisor, which determines when and how to execute the requested adjustment. The vehicle remains autonomous throughout the experience. The simulator also records passenger interventions, demonstrating how this information could eventually contribute to improving autonomous-driving behavior.

How I built it: I built TheFrontierMile using Next.js, React, and TypeScript, with OpenAI Codex assisting throughout development.

The application combines a vehicle cockpit interface, an interactive touchscreen, and a scripted autonomous-driving simulation. Rather than creating three disconnected animations, I developed the scenarios around shared simulation state so that vehicle speed, lane position, surrounding traffic, and driving decisions remain consistent across the visualizations. The simulation includes a supervisory decision layer that processes passenger requests, evaluates scenario-specific conditions, and determines whether a maneuver should proceed, be deferred, or be rejected. I also incorporated automated testing to verify the driving logic, intervention behavior, and continuity of the autonomous-driving session. GitHub was used for version control and milestone management throughout development.

Challenges we ran into:My biggest challenge was translating the vision in my head into a functional simulation.

I didn't want TheFrontierMile to be just an attractive touchscreen with buttons that displayed predetermined responses. I wanted passengers to see the vehicle driving, understand why they might intervene, and observe how their input actually changes the simulation. Achieving that required several iterations of the visual environment, vehicle positioning, driving logic, and interaction design. The highway scenario was particularly challenging. It required accurately representing three lanes, a slower lead vehicle, and a semi-truck in the far-left lane while demonstrating the difference between the vehicle's original lane-change plan and the passenger's preferred timing. Another challenge was improving the simulation without sacrificing the interactive touchscreen or the cockpit experience I had already built. Through iterative prompting, testing, and visual feedback, I was able to bring the implementation closer to the original concept.

Accomplishments that we're proud of: I'm especially proud of how much TheFrontierMile evolved throughout development.

The project began as a basic 2D driving simulation. With the help of AI-assisted development, I transformed it into a more immersive experience featuring a vehicle cockpit, an interactive touchscreen, and driving scenarios that demonstrate the relationship between passenger input and autonomous behavior. I'm also proud that the project goes beyond simply presenting an idea. The simulation allows me to demonstrate how a passenger's request can change a planned maneuver while autonomous driving continues. On a personal level, I became much more comfortable using GitHub, managing repositories, committing milestones, and working through an iterative software-development process. As a solo builder, seeing the project progress from an initial concept to a functioning application has been one of the most rewarding parts of the hackathon.

What we learned: One of the most important things I learned was the role of simulation in autonomous-vehicle development.

Testing new driving behaviors directly on public roads can introduce significant risks. Simulation provides a controlled environment for exploring scenarios, evaluating decisions, and observing how a system responds to different conditions before considering real-world testing. I also learned how important it is to distinguish between a technically permissible driving maneuver and a passenger's personal preference. Autonomous driving is not only about reaching a destination. The passenger experience matters, too. From a development perspective, I learned how to translate natural-language ideas into specific system requirements, use AI-assisted coding tools, validate behavior through testing, and refine a product through repeated visual and functional iterations.

What's next for TheFrontierMile: My long-term vision is for TheFrontierMile to become more than a passenger-control interface.

I see PAADD as a potential bridge between human driving preferences and autonomous-driving intelligence. In the short term, passengers could use it to communicate contextual preferences without unnecessarily disengaging autonomous driving. Over time, aggregated intervention data could help researchers and developers identify recurring situations where passengers prefer different driving behavior. That information could potentially inform future improvements to autonomous-driving systems, allowing them to better account for human context and passenger comfort. Ultimately, the goal is for this type of interface to become less necessary as autonomous-driving systems learn to anticipate these preferences themselves.

Built With

• codex
• nextjs
• react
• typescript