Beyond Navigation

Beyond Navigation Logo

Inspiration💡

Tired of mall mazes and endless searches for the elusive restroom? Beyond Navigation is here to save the day! Inspired by the comical struggles of our team member in a race against time with nature, we've crafted a out-of-the-box solution for internal navigation, no GPS required. Say goodbye to wandering and hello to efficiency! #HackTheMaze #BeyondNavigation

What it does 🔎

Inspired by the convenience of navigation kiosks, our solution tries to bring a similar kind of functionality without the need for GPS, dedicated kiosks or expensive internal map technology. Here are the key points on how our whacky solution tackles the maze problem:

Leverages unnoticed evacuation / floor plans mandated in most buildings to generate a map outline of the entire floor
Users can simply type in their desired destination and the chatbot generates the best path to reach the location
Unlike traditional apps cluttering your phone, our solution is a WhatsApp chatbot, making is cross-platform on both Android and iOS devices and ensures instant, hassle-free access to internal navigation, right within your familiar WhatsApp interface

How we built it 🔨

WhatsApp + Infobip API - We adopted to go for a conversational UI via Whatsapp, improving reach across mobile platforms and eliminating the need for additional app downloads.
OpenCV + pathfinding algorithms- Determine general layout of the floor from the image and to calculate the shortest path.
Google Cloud Storage - Secure storage of user images as well as computed data.
Google Cloud Vision AI API - Used to identify rooms in the floor plan and determine their positions accurately. This integration allows the user to dynamically query the map
Google Cloud Run Platform - A low-cost serverless computing environment where we run our backend

Challenges we ran into⚠️

OpenCV itself: Processing images was quite a challenge, especially when we used live images from WhatsApp camera. The quality of images varied, making the processing part pretty tricky.
New technologies: We want to challenge ourselves with choosing technologies out of our comfort zone. Most of the technologies that we have used in this project were something new to us
Comprehensive Google Cloud Usage: Hosting the whole solution on Google Cloud was a first for us. Learning to navigate through the quirks of Google Cloud became a crucial part of our hackathon journey

Accomplishments that we're proud of🥇

Complete prototype solution: We built a prototype that actually works! It's not just an idea; it's somthing ready to help you find your way around
Smart Tech Choices, No Big Costs: LLM models is a versatile tool, capable of even ingesting however, it is extremely expensive to run, especially at scale. We didn't go for fancy and expensive tech that might sound cool. Instead, we chose practical and budget-friendly technologies.
Optimizations: For real-time messaging, we needed really quick response times. Initially, our application took 45 seconds to process the pathfinding algorithm which was unacceptable. We dug deep and found a significantly more efficient albeit slightly less accurate Python library which did the same in 2-3 seconds bringing our overall processing time to ~7 seconds.

What we learned 🧠

Dealing with user input: We learned a lot about how unpredictable user input as well as intermediate systems can be (WhatsApp image compression looking at you 👀)
Image Manipulation: We learnt new ways to interact with images, the pipeline used to extract useful information from them as wells as the nuances on working with them

What's next for Beyond Navigation 💭

Although the idea originated from the nuisance of new surroundings, we discovered that this idea can greatly improve universal accessibility by including features that can create a personalized path to accommodate disability constraints to optimize route planning. To improve the accuracy of our pathways, we plan on implementing Bluetooth and Wifi. Additionally, we wanted to use the phone’s magnetometer and pedometer to enhance user guidance precision.