Cognitive Automotive Racer

Optimization algorithms and models for autonomous cars in a racing scenario.

Overview

Most racing car scenarios are built with manual racing controls. This project is aimed at autonomous racing using advanced Machine Learning and Deep Learning algorithms.

Racing cars have always tested the limits of our understanding of path optimizations along with acceleration and declaration of the vehicle on a winding track. Once multiple racing cars are introduced the driver of the vehicle must not only optimize the path but avoid any obstacles (cars) in the path. This makes it a very complex problem to solve for an autonomous vehicle. This project aims to take up this challenge in a phased approach.

Architecture

The diagram illustrates the intercommunication between the components. For details please refer to the components.

Personas

Designing a solution requires empathy for the users. Here are some of the personas who may benefit from the product.

Racing
Team

Team is responsible for the car model and all of the mechanical, electrical and software components.

Race
Spectators

Spectators can access viewer data directly from the proximity network or from the Cloud.

Race
Administrators

Race administrators organize and monitor individual races. They support and on-premise racing infrastructure.

Components

This section shows all the major components of the project. The components are grouped under Hills or Epics for Cognitive, Internet-of-Things, etc. and Foundation.

Race
Car Model

Tamiya TT02 or similar scaled model for the car with custom fitted GPU and ARM based single board computer, cameras and sensors.

Data
Collector

On-premise network for collection of data from multiple vehicles and forwarding the information reliably to the Cloud.

Self-driving
Model

Reinforcement Learning model for accepting multiple inputs, algorithm for lane detection for inference on the car and simulation and training environments on the Cloud.

Projects

Collaboration with universities IBM Center for Advanced Studies and IBM Academy of Technology. Each Open Project is an iteration of a component of the asset.

North Carolina State University

Jeremy Edmondson,
Associate Director, Senior Design
Electrical and Computer Engineering

Project Id:

car-cog-cloud-002-ncsu

Program:

Component:

Groups:

Students:

North Carolina State University

Dr. Rachana Gupta,
Director, Senior Design
Electrical and Computer Engineering

Project Id:

car-cog-onprem-002-ncsu

Program:

Component:

Groups:

Students:

IBM Academy of Technology

Arunava Majumdar,
Lead, CAS Chicago, US-MW, India,
Lead, IBM Asset Strategy,
IBM Cloud Platform

Project Id:

car-phase-ii

Program:

Component:

Groups:

Developers:

“I always thought records were there to be broken.”

Michael Schumacher

Blog at WordPress.com.