Automated Driving System and ThingSpeak Seminar

Data Visualizaiton Example on MATLAB Automated Driving System Toolbox

Here’s an account of the Automated Driving System and ThingSpeak Seminar by Mathworks that I attended on Friday, September 14, 2018.

The seminar began with a talk about deep neural networks and Matlab’s ability to natively support Caffe, AlexNet and other widely used models – without any setup, right out of the box – which is amazing! I have spent several hours to put a system together and each time I must change my working system, I need to go through that painful process again. I can see value in the time saved by enterprises when they get to development faster by skipping the setup altogether and deploying the MATLAB toolbox instead. The ADS toolbox also supports the Nvidia NGC which will enable the users with access to superior graphics processing in the cloud.

Another feature that caught my eye was the ONNX support. ONNX is an acronym for Open Neural Network Exchange, which is a format to represent deep learning models. This bridge to the open source community is an essential feature. This will enable users to export and import deep learning models to and from the open source community. ROS is also supported by Robotics System toolbox which was released earlier.

I was certainly impressed by the new live-editor which has some similar features as the Jupyter Notebok. I think this feature creates a great value for large organizations for knowledge retention when there is a change in personnel. The live-editor lets the user attach text, media, links and even equations to the code along with the ability to run the code in sections. This can make not only debugging but also documentation very easy and would mean the engineers get to spend more time on solving the problem than on the documentation.

Most engineers agree that live tech demos are tough. The live data visualization and simulation features that Arvind managed to demo were very neat. I love rviz and gazebo for robots where having detailed URDF models is indeed helpful. For autonomous cars, there is no need for details and MATLAB simplifies this by using cuboids for vehicle simulation. I think using gazebo to simulate for all the different models of cars is counter-intuitive. However, MATLAB does offer 3D visualization of the road and vehicle with the unreal engine.

I learnt a lot about NXP’s Blue box as well as the ThingSpeak platform for IoT applications. It was a good way to spend a Friday morning.


Leave a Reply

Your email address will not be published. Required fields are marked *