Machine Learning and Deep Learning in Robots

Recently, the FSL Year 2 Engineering Stream has been working on building robotic vehicles made from 3D printed parts and circuitry in conjunction with two microbit controllers. As of current, these robots utilize the positioning of the microbit controller to determine whether to go forward, backward or to stop. For example, if the user tilts the controller 45 degrees away from themselves, the robot will move forward.  Working on this project made me extremely curious about how we can implement Machine Learning in our projects. For many, we hear terms such as Machine Learning and Deep Learning from a plethora of conversations, but it may be hard for some to truly understand what these concepts mean.
Mike Mackenzie © CC BY-SA
With many progressive advancements in the field of Artificial Intelligence, it is important to note that these advancements can be classified as either machine learning or deep learning. Machine Learning is a methodology of data analysis that automatically works on analytical model building. The big idea in this branch of AI is that it is possible for systems to analyze data, look at patterns based on the data as well as make decisions without much human support. Now you may be wondering, how is this phenomenon used in our daily lives? Machine Learning is a critical part of our everyday lives from online recommendation systems such as YouTube, Apple Music, Netflix, Udemy etc. to credit card fraud detection which has helped secure transactions. Part of the reason why companies and researchers are using Machine Learning in their products and services is their ability to quickly recognize patterns of data to benefit companies. You may be wondering what would happen to these algorithms if they came across more complicated datasets, and how a computer is able to replicate human abilities such as recognizing speech (i.e. Siri), identifying images (i.e. Facebook) or even making predictions (i.e. Stock Market Prediction Algorithms). The answer to this question is called Deep Learning, which essentially is a branch of Machine Learning that is able to do the tasks mentioned above. Deep Learning is able to do these tasks by setting up parameters about the data and through training the computer to do this on its own using many layers of processing. The algorithms associated with Deep Learning has allowed machines to classify, detect, recognize, describe and understand the data inputted into its systems with much more detailed analysis. Now to understand the potential machine learning and deep learning has on robots. Take something like Google’s Self-Driving Car. The Self-Driving Car, as the name implies is a car that autonomously drives itself around without much user intervention. As stated in the article, there is “no steering wheel….simply a start button and a big red emergency stop button.” The car works by using a frame of a modified Toyota Prius coupled with sensors and machine learning software to recognize people, other cars, road signals, street signs, objects and even unpredictable hazards such as cyclists or oncoming accidents. Some other applications of Machine Learning in Robotics include: Imitation Learning (a behavior of imitation also exhibited by rather inquisitive toddlers), self – supervised learning robots (such as the Google Smart Car’s Road Detection Algorithm) and Assistive and Medical Technologies such as Baymax from Disney’s Big Hero 6, a robot that is able to assess a patient’s needs and supplement them accordingly. Now with the understanding of Machine Learning and Deep Learning, I hope to install these algorithms in my FSL Robotics Project.

Speak Your Mind