UbiR | Devpost

UbiR: Ubiquitous Robotaxis
Our Simulation Tool
Customer Overview
Fleet Overview
Overview Dashboard
Live Map

Inspiration

First of all, it was important for us to understand the purpose of the challenge, in order to orient our solution appropriately. We then evaluated existing solutions and built upon them, adding our creativity and technical skills.

What it does

Our fleet management system comprises two major parts:

A powerful commissioning algorithm that allows great reductions in:
- A customer waiting times.
- Distance traveled by the robotaxis.
A dashboard that serves as a fully-featured control tower for the robotaxi fleet, offering:
- Real-time view of the vehicles.
- Useful statistics and metrics.
- An integrated simulation engine (SCENE) to assist with decision-making.

How we built it

For the commissioning algorithm, we employed a genetic algorithm (GE). The idea behind a genetic algorithm is very similar to real genetics: there is a population of individuals, each of them with their own genotype. These individuals mate, producing an offspring that combine genes from both successors. Moreover, mutations are also possible. The algorithm keeps track of all the different populations as they evolve, keeping track of the "best" populations, i.e., those that are the closest to our objectives. In our case, the individual making up a population are the taxi-customer assignments, and the objectives we optimize for are customer waiting time, and distance traveled by the robotaxis.

For the front-end we employed React Native and Grafana, and our backends are powered by Flask, PostgreSQL and Dockerized into microservices.

Challenges we ran into

There were plenty, but the biggest ones were:

Adapting the genetic algorithm to our specific problem.
Coordinating the multiple pieces that make up our system (commissioning algorithm, simulator, backend, frontend, etc.).
Designing a sleek UI/UX for our frontend.