Image course: https://www.pinterest.com/pin/264938390560462730/

“Autonomy: The Quest to Build the Driverless Car” is a story of the self-driving industry. Lawrence D. Burns, the author of the book, is a former General Motors executive and an advisor to Waymo.

The book is full of fascinating stories about the evolution of a crazy idea of self-driving cars from a handful of small teams to the current state of the industry. Burns was a first-hand participant in many events he describes.

“Talks at Google” with Lawrence Burns

He starts the story from DARPA Grand Challenge, a competition for robocars, which would be held on March 13, 2004…


Beginning of July. It’s time to look back at the first half of the year. Reviews of previous years were here: 2016 and 2017.

Writing

Year to date, I wrote ten blog posts, it’s three more than for the first half of 2017. The most popular post written in 2018 is “ResNet for Traffic Sign Classification With PyTorch.”

Self-driving Cars

As you could notice, I’m writing mostly about deep learning and self-driving cars. Self-driving cars is an interesting field with numerous unsolved technical problems. I have been passionate about building autonomous systems since childhood. In June, I joined Drive.ai, a startup that uses…


High definition map. Image source: https://arstechnica.com/cars/2017/03/the-most-detailed-maps-of-the-world-will-be-for-cars-not-humans/

I used to think that the maps are solved. We have great maps, primarily Google Maps, where roads, buildings, bridges and all other interesting physical objects are mapped. I thought about the future of maps is a series of incremental improvements and ongoing maintenance. Until I learned about HD maps.

Regular maps are not sufficient for autonomous vehicles. They don’t have information about traffic lanes, traffic signs, and lights, position, and height of the curbs. Also, they are useless for localization purposes. They are designed for humans, not for machines. …


Image source: https://www.echelonfront.com/extremeownership

There are plenty of books about management. Rarely you can find a book that’s worth reading. “Extreme Ownership: How U.S. Navy SEALs Lead and Win” by Jocko Willinck and Leif Babin is an exception. It is written well and describes simple actionable principles of management with examples.

Jocko and Leif were leaders of SEAL Team Three’s Task Unit Bruiser. They learned valuable lessons about leadership in a tough environment: Ramadi, the most violent and dangerous battlefield in Iraq. When they returned from deployment, they were training the next generation of SEAL leaders. …


Image source: https://www.nasa.gov/mission_pages/apollo/apollo11.html

Kalman filter is an algorithm that combines information about the state of a system using predictions based on a physical model and noisy measurements. It is called a “filter” because it is filtering out measurement noise.

Kalman filter has many applications in robotics. For example, it can be used for localization of coordinates and velocity of a robot based on measurements of distance to certain landmarks. It’s widely used for filtering information from GPS sensors and radars. Kalman filter was in Apollo navigation computer that took humans to the Moon and back.

Kalman filter is unimodal. If you need to…


Image credit: https://c2.staticflickr.com/4/3367/3630171023_f56b62760b_b.jpg

Let’s talk about transferring ideas. From robotics. To software development management. Should be fun, right?

Some time ago I was in a team meeting where the team was discussing and estimating User Stories. User Story in agile software development processes is an atomic independently deliverable feature. Something that is small enough, typically just a few days of development, and yet somewhat useful for the end user. Estimation process starts with somebody, typically a person that created the User Story, describing the requirements. It is followed by a discussion about why the hell we really need to implement the feature, how…


We use the phrase “rocket science” to denote a system or activity that is extremely complex. As history shows, it is very hard to design and build a vehicle that accelerates to 7.8 km/s (orbital velocity needed to maintain a stable low Earth orbit) and have precise control on its trajectory. Anatoly Levenchuk argues that it is much harder to control a vehicle that moves safely on real roads with 300 times smaller speed, 25 m/s (90km/h), near other vehicles with similar speeds. …


German Traffic Sign Recognition Benchmark dataset is probably the most popular image classification related to self-driving cars. Autonomous vehicles need to detect and classify traffic signs to understand traffic rules applied to a segment of the road. Perhaps, this dataset is too small and incomplete to use it for real applications. Nevertheless, it is a good benchmark of computer vision algorithms.

Dataset

The dataset consists of two parts: a training set and a test set. The training set contains 39209 images of traffic signs classified into 43 classes, such as stop sign, bicycles crossing, and speed limit 30 km/h.


Image credit: FleetOwner

Self-driving cars with remote human operators could be tested in California starting from April, VentureBeat reports. It’s not certain. The new regulations need to be approved by California’s legal compliance agency. Then the DMV will open a 30-day public notice starting from March 1.

A few companies, such as Nissan, Waymo, Zoox, Phantom Auto and Starsky Robotics, have been working on remote control technology. It is an interesting strategy that allows using autonomous vehicles in more use cases sooner. Vehicles can drive autonomously in simpler situations, and a human takes control when the autopilot can’t operate safely.

It seems to…


Photo by Jared Erondu on Unsplash

Learning opportunities are endless these days. This month I visited “Age of AI,” a small conference in San Francisco. The conference explored the current state and the future of AI on different levels. On a micro level, there were a few talks about recent breakthroughs in reinforcement learning and generative adversarial networks. On a macro level, there were good talks on founding new AI startups, the impact of quantum computing, and, of course, a great talk about future of the AI by Tim Urban.

Pavel Surmenok

Machine learning engineering and self-driving cars. Opinions expressed are solely my own and do not express the views or opinions of my employer.

Get the Medium app

A button that says 'Download on the App Store', and if clicked it will lead you to the iOS App store
A button that says 'Get it on, Google Play', and if clicked it will lead you to the Google Play store