This Week In Autonomous Vehicles — August 27th Edition

We win by shipping/news.voyage.auto

This week has been a very busy one for me personally and the rest of the autonomous vehicles’ industry. Its another weekend when I give you the latest developments in the autonomous vehicles’ industry. So this week I have had to make some painful decisions which will see me not being able to graduate (2019) as expected, but all the same, life has to move on. I will still work hard to become a self-driving car engineer. In some good news, an intern at Voyage is shipping already:-). In Kenya, the National Transport and Safety Authority has urged drivers to use dash-cams to record whats happening on the road, a good move which will make them cameras cheaper to buy and also provide raw data for training my neural network on lane detection.

During my free time I read about 5 questions that I should ask myself daily to achieve success. This is an article that I highly recommend to all my readers. Across the world in US, Elon Musk’s ambitious goals for Autopilot technology have prompted safety warnings and resignations. I hope this is sorted soon so that I can get myself a Model S soon. At Udacity they are always coming up with something and this week was no different, they have launched their newest Nanodegree program: Data Foundations! That’s enough for a summary, lets get down to the articles this week.

I’ve gotten to build a very, very real feature into the coolest product out there, a self-driving taxi — Michael Gump

Michael Gump is a Deep Learining Intern at Voyage (Self Driving Taxi Company) and he has contributed to the autonomous driving tech with his new product. Here is an excerpt from news.voyage.auto

This summer I’ve been working on several problems in computer vision, which means reading a lot of papers and training a lot of models. Most generally, I’ve been working with public datasets to classify objects in LIDAR data. When we as humans see the world, we automatically perceive depthand produce information about objects like “that’s a car”, or “oh no, that’s a person in the middle of the highway, hit the brakes!” Most of my work this summer has been on bringing these types of insights to Voyage’s self-driving taxi, using a 3D view of the world (LIDAR point cloud) and deep learning.

Black & White image constructed from a LIDAR observation (no class information), then with added class information (green: human, blue: car)

One of the first things I shipped this summer was a good, quick visualizer. Using Vispy let me visualize large point clouds in sequence and debug my models in conditions as similar to the real world as possible. Another lesson from this summer is that a good visualizer should alwaysbe the first step for problems like this. It can be very difficult to figure out what the problem is from just the model’s loss curve.

To build a set of tools for training and validating my models I used PyTorch. I wasn’t familiar with PyTorch before but it’s become my new favorite framework. It loses some of the specificity of TensorFlow but it is much easier to get started.

Encoder-Decoder Network in action. The model’s raw, noisy predictions fade into the extracted real object labels.

One of the models I built was an Encoder-Decoder network that transforms several channels of input data into categorical predictions. From these noisy predictions we can infer the real categories of all the objects in front of us. This kind of model is powerful because it can become independent of certain sensor and processing errors. For example, a model relying on the size and shape of objects to predict their class can be prone to detection errors. An encoder-decoder model can sidestep issues like this by recognizing patterns in the scene and turning them into predictions directly.

Headlines from Google, Uber, and Waymo

David Silver has done a great summary of what Google, Uber and Waymo have been upto this week. These 3 companies are always making headlines with their products as each hopes to be the first in this field.

The first world is Waymo’s physical testing facility at the old Castle Air Force Base, in California’s central valley. The article talks about a city with streets but no buildings, designed specifically for testing self-driving cars. When Waymo runs into a particularly sticky driving situation, they just pave a version of the streets on their test facility and run their cars through that scenario over and over and over again.

You can read the full article here: https://medium.com/self-driving-cars/headlines-from-google-uber-and-waymo-d4bee7325291

Ustwo — Humanaizing Autonomy

Rick Fish, 41

The team over at Ustwo, a company that works with other companies to create amazing digital experiences, has published and is almost ready with their second book called Humanaizing Autonomy . Here is an excerpt…

Our first book made industry predictions, suggested user experience (UX) solutions and offered working practices to help with in-car UX problems. And we’ve been pleased to see some of these adopted in current vehicles. Alfa, for example, returned to a haptic controller in their Giulia — partly because this is the class standard but you could argue that it’s rectifying the “slap a touchscreen on it” approach we discussed in our first book.

More recently, we’ve been working on another book — Humanising Autonomy: Where are we going? This time we’re turning our human centred design (HCD) approach to the barriers to adoption facing autonomous vehicles (AVs) and focusing on the new experience challenges that come with it. How can the magic of technology combine with a deeper understanding of how we behave to make the actual experience better and truly adopted by all?

Here we’re sharing some of what’s to come in Humanising Autonomy before its full release later this year. This book is about the new mindset that’s needed in order to design the experiences of the future. It’s about questions. And about People.

You can read the full article here:https://medium.com/ustwo/what-does-autonomy-mean-to-you-987f1f60c458

Playing In The Sticks: Self-Driving Cars And Rural America

Blue: self-driving / Green: connected / Brown: anything goes

This article by Mitch Turck looks at Self Driving Cars from a rural America’s perspecive. We are in the race to develop autonomous vehicles for urban areas but what about rural areas? How will they fair out there in the dusty no sign roads? Here is an excerpt…

Once you hit publicly funded local roads (in green), you start to become a liability and responsibility to others… even those folks far off in the city, whose taxes help to cover infrastructure costs if you wrap your truck around a pole. For the sake of all the autonomous traffic moving through and around these areas, any vehicle on these green roads should be connected. Not necessarily the “gubment gone and took my privacy away” connected though — just the kind of anonymous feedback signal that helps the grid understand where potential obstacles might arise. We’re talking about transponders that cost virtually nothing and reveal virtually nothing; this should be an easy sell for anyone who doesn’t have Don’t Tread On Me tattooed across their entire face. So, if you want to ride your moped from the backwoods to the gas station, your choices would be

Read the full article here: https://medium.com/@mitchturck/playing-in-the-sticks-self-driving-cars-and-rural-america-b1c89554bcba

Thats all that has been buzzing in the world of Autonomous cars, if you write about Autonomous vehicles and I have not included your article, do give me a shout out and I will add it in.

Otherwise, thanks for reading, please give this article some claps there on the side, thanks and take care when testing your autonomous vehicle.

[circuit open]

)
Fundi Willy 楚大洋

Written by

Electrical, Automation & Self-Driving Vehicles | Udacity Self Driving Car ND | My Garage >> www.githuka.com

Welcome to a place where words matter. On Medium, smart voices and original ideas take center stage - with no ads in sight. Watch
Follow all the topics you care about, and we’ll deliver the best stories for you to your homepage and inbox. Explore
Get unlimited access to the best stories on Medium — and support writers while you’re at it. Just $5/month. Upgrade