If you fancy building your very own four-legged quadruped robot you might be interested in a project published by Samira Peiris to the Hackster.io website. Using a Raspberry Pi 2 Model B, Samira Peiris has created a small programmable robot and equipped it with deep learning technology thanks to TensorFlow, an open-source software library for machine intelligence. The Robot comes in a kit form and can also be controlled by an Arduino Zero development board if preferred.
“TensorFlow is an open source software library for high performance numerical computation. Its flexible architecture allows easy deployment of computation across a variety of platforms (CPUs, GPUs, TPUs), and from desktops to clusters of servers to mobile and edge devices. Originally developed by researchers and engineers from the Google Brain team within Google’s AI organization, it comes with strong support for machine learning and deep learning and the flexible numerical computation core is used across many other scientific domains.” For more resources and information on the open source TensorFlow machine learning project and application jump over to the official website by following the link below.
Samira explains more about his conclusions about the build and the implementation of Keras and TensorFlow. “The LeNet model is fairly easy to implement using Keras and TensorFlow, and depending on your dataset, it can produce very accurate results. This code could be improved so that the quadruped can correct its alignment with the white line, but this is quite difficult to do with the creep gait. (Since the turning angles are fixed and considerably large). You could implement better gaits by making them from scratch, or by using ROS. Here’s a really good case study about building legged robots using the Robotic Operating System.”