Thanks to the collision of machine learning and the Internet of Things (IoT), driverless technology is advancing rapidly, and advancements are expected to continue on this trajectory with technology giants like Google, Intel, Mercedes and Tesla all working on creating the smartest vehicle on the market.
Many companies working in this area are pledging fully-autonomous cars to be available in less than 5 years, which is sure to have a huge impact across business and society, and could be revolutionary in industries such as transport.
Teymur Sadikhov is Senior Vehicle Intelligence Engineer in Autonomous Driving at Mercedes-Benz R&D, where he works on many aspects of autonomous vehicles; from developing decision-making and behaviour architecture in self-driving cars, to designing their trajectory planning algorithms and adaptive controls.
I asked him a few questions ahead of his presentation at the Machine Intelligence in Autonomous Vehicles Summit this month, to learn more about recent advancements and key challenges in autonomous vehicle technologies, and what the future holds for the field.
What started your work in autonomous vehicles?
I’ve been working on control and robotics for different kinds of systems, from single agent to multi agent, from theoretical control to applied multi agent robotics, for the last 9 years. Having an aerospace engineering background I worked on trajectory planning for autonomous helicopters at United Technologies Research Center and that together with my passion for autonomous driving led to my work at NASA Jet Propulsion Lab, where I worked on localization, control and multi agent behaviour planning for autonomous ground vehicles. I continued to my passion in my current job.
What are the key factors that have enabled recent advancements in autonomous vehicles?
The key factors for recent advancement in autonomous vehicles in general, including aerial and ground vehicles, were enabled by the recent advancements in robotics and machine learning (ML). Moreover, sensors and computing power getting cheaper had a major effect on the advancements as well, because it enabled real time sensing and planning for these kinds of systems using only on-board computing power. If we consider autonomous ground vehicles in particular, I’d say the advancements were motivated and powered by DARPA (Defense Advanced Research Projects Agency) Grand Challenge (2004, 2005) and Urban Challenge (2007), where autonomous research vehicles from many universities and companies from around the world competed in autonomous driving competitions. These events gave rise to many projects including Google’s autonomous driving project.
What are the key challenges to progressing autonomous vehicles?
One of the key challenges for autonomous driving is driving in urban environments since it presents a lot of uncertainties and a lot of moving objects which in turn create challenges for decision making. I’d say highway driving is an almost solved problem and more development nowadays focuses on urban driving. Moreover, computing power and sensor cost still create issues for bringing the technology to the market. Therefore, there are a lot of companies which try to create cost effective sensors, especially LIDARs.
Another challenge would be regulations as there are no complete set of regulations related to autonomous driving on public roads. This point actually holds true for autonomous aerial vehicles as well since FAA hasn’t approved complete sets of regulations for them either. I’d say human drivers create another issue for autonomous driving as it’s not always easy to predict a human drivers’ behaviour. If all vehicles were autonomous the problem would be simplified.
Another challenge in this area would be security against malicious attacks. From an ML perspective, the amount of collected data can create challenges as well.
What developments can we expect to see in autonomous vehicles in the next 5 years?
I believe we’ll see a lot of developments in autonomous vehicles in the next years, including advancements in robotics and ML, cheaper sensors and more GPU power. Also, regulations hopefully will be set in place. We’ll see fully automated driving on highways and platooning autonomous driving trailer trucks. However, to predict that urban driving will be fully autonomous and legally allowed might be a little early.
Outside of your own field, what area of machine learning advancements excites you most?
I’m very excited about the advancements in ML in general and deep learning in particular. For example, IBM Watson and Google DeepMind’s recent advancements are very promising. I’m especially excited about the fact that ML is being used to go through enormous amounts of medical research data and create personalized and new treatment options for cancer patients. I believe one day it will help us eradicate cancer altogether and hopefully that day will come sooner than later.
Teymur Sadikhov will be speaking at the Machine Intelligence in Autonomous Vehicles Summit, taking place alongside the Machine Intelligence Summit, in San Francisco on 23–24 March. Meet with and learn from leading experts in autonomous vehicles, IoT, the smart dashboard, machine learning methods and predictive analytics. Register to attend here.
Other confirmed speakers include Sam Kherat, Sr Manufacturing Automation Team Leader, Caterpillar; Gary Marcus Director, AI Labs at Uber; Luca Rigazio, Director of Engineering at Panasonic SV Lab; and Pratik Brahma, Machine Learning R&D at Audi/VW. View more speakers and topics here.
See the full events list here for summits and dinners focused on AI, Deep Learning and Machine Intelligence taking place in San Francisco, London, Amsterdam, San Francisco, Boston, New York, Singapore, Hong Kong, and Montreal!