The problem ZenRoute solves

"Hey Maps, I've got time. Let's take the scenic route, show me some peaceful places!" Maps replies, "Sorry, I'm not programmed for that." Zenroute steps in with, "Hey there! I've got you covered."

We're tired of the rush of modern travel in search of a serene journey that lets us immerse in nature's beauty.

The Problem:

In a world where haste overshadows the joy of the journey, travelers yearn for a peaceful, aesthetically pleasing route. Traditional mapping apps prioritize the quickest route, neglecting the need for a peaceful experience.

The Solution:

ZenRoute helps travelers find serene, less-traveled routes. Whether you're a road tripper, cyclist, or scenic route enthusiast, Zenroute's AI-driven chat feature, Zorro, is your reliable guide for navigation, support, and solutions.

How Zenroute works:

Personalized Zen: ZenRoute presents multiple scenic route options tailored to your preferences based on your origin and destination, including lakes, viewpoints, dining spots, and worship places.

To use ZenRoute, simply provide your source and destination and instruct Zorro. For example, say, "Guide me to my destination, including several lake stops along the way, and ensure the journey doesn't exceed 3 hours. Also, I have my dog with me." Zorro will offer suitable routes with a single click.

The process starts with the user's intent-based instructions. The Language Model (LLM) extracts relevant details. Extracted data is converted to JSON and fed into a modified Dijkstra algorithm. This algorithm considers source, destination, and time constraints, retrieving feel-good places from our database. It aims to maximize feel-good stops while meeting time constraints.

AI filters and identifies the top three routes based on additional user-provided constraints (e.g., traveling with a pet dog).
After selecting a route, ZenRoute provides navigation guidance using the MapBox API. User trip data is stored to improve recommendations over time.

Challenges we ran into

One of the primary hurdles we encountered was the development of an effective Dijkstra algorithm. Ensuring its accuracy and efficiency in processing user input, extracting essential properties, and computing the ideal route with maximum feel-good places was a critical challenge. We had to carefully consider factors like route optimization, time constraints, and the multitude of feel-good places along the journey to ensure our algorithm delivered the best results.

Occasionally, the Language Model (LLM) produced incorrect results in understanding user intent. To address this, we relied on swift and clear communication to clarify user requirements.