BETA
This is a BETA experience. You may opt-out by clicking here

More From Forbes

Edit Story

Open Source AVs: The Story Of AV Development In Estonia

Following
This article is more than 3 years old.

The article “Linux beat IBM will open-source software beat Waymo and Tesla ?”  introduced the power of the open-source organizational model to crowdsource the innovation of a community.

Is it possible to build a functioning Autonomous Vehicle (AV) Shuttle with a small team by using open-source methods ?

A research team from the Tallinn University of Technology (TalTech) in Estonia took on this challenge, and the result was the ISEAUTO shuttle shown in the picture above operating in snowy conditions ! We explore their journey in engaging with open source AV technology. 

Disclosure: The TalTech and the Advanced Mobility Institute at Florida Polytechnic University have a collaboration agreement. I initiated and built the Advanced Mobility Institute at Florida Polytechnic University.

TalTech is a leading polytechnic university in Estonia. In the world of technology, the area is best known for providing some of the key founders of Skype. Like many AV design teams, TalTech’s journey towards building an AV started with robotics competitions such as Robotex. In June 2017, an ambitious self-driving vehicle project was started with the goal to develop a low speed AV Shuttle for the university anniversary in September 2018.  

"We invested in the project because we believed that developing AV technology was such a multifaceted challenge that every ambitious academic community at the university could contribute. We thought that TalTech had both smart people and novel ideas to participate," said TalTech rector Prof. Jaak Aaviksoo. 

With millions to billions of dollars being spent on AV development at companies such as Waymo, is such a task possible?

The project was a joint effort between Taltech and Silberauto Estonia, a distributor of automobiles. Silberauto focused on body design and mechanical assembly. Meanwhile, the TalTech team had the daunting task of providing the whole AV electronics and software subsystem. The team consisted of approximately 18 students working with dedicated faculty.  Overall, a relatively small team which had to implement all the major functionality in an autonomous vehicle as shown in the figure below.

Due to time and cost constraints, the hardware components used were off-the-shelf (Velodyne LiDar, Standard PC with Nvidia graphics cards, Point Grey Cameras, etc). The key daunting challenge was the software system which had to be customized for the application and location. The only viable path was to start with a base consisting of open-source environments.  However, which one?

An investigation (link to detailed analysis) of open-source environments revealed four major candidates — Nvidia Drive , OpenPIlotBaidu Apollo, and Autoware. Drive, OpenPilot, and Apollo were sponsored by commercial entities. Autoware was sponsored by a research group from Nagoya University in Japan.

When one chooses an open-source environment, the evaluation is not only the current functionality, but other important factors such as likely momentum of the community. In TalTech’s evaluation, Autoware was chosen based on completeness, independence (from particular automobiles or electronics hardware), and governance structure(company led vs independent foundation). 

Using the open source strategy and with a small student team, the team built a solution in time for the Taltech anniversary. The performance characteristics were in line with other last mile solutions as shown in figure above (reference). Uniquely, the team was even able to demonstrate the operation of the shuttle in snow (Video). Also, after the prototype launch, the demand for commercialization was very strong, so a company called AuVeTech has been built for that purpose.

The immediate takeaway:

To build an autonomous vehicle is not an impossible task and something only big OEM can do. A key to the success was the availability of open-source software upon which we could build our system design,” said Raivo Sell, Senior Researcher and Program Manager of Product Development and Robotics Company at TalTech University.

Building a research prototype with students is a reasonably common story. As Stefan Seltz-Axmacher, the CEO of Starsky Robotics, points out in his excellent medium article, building something that “works” is an accomplishment. However, building a true product which is statistically reliable is 10-1000X more work.  

The unique circumstances of the Taltech ISAUTO project are that it has now created an open-source platform with a working shuttle. This is one of the primary reasons that the Taltech project attracted the attention of the Advanced Mobility Institute a continent away at Florida Polytechnic University.

Can the 10X-1000X cost curve be bent with the collective innovation engine of the open source paradigm ?   

We are about to find out in the next few years. 

Note: For those who are interested, the story of Autoware is available at “Will A Small Open-Source Effort From Japan Disrupt The Autonomous Space ?”

Follow me on Twitter or LinkedInCheck out my website