140 – Self-Driving Cars
This episode is about autonomous vehicles (aka self-driving cars). Our guest is Jonathan Sprinkle, assistant professor at the department of Electrical and Computer Engineering at the University of Arizona. We talk about the topic mainly from the perspective of (software and systems) engineering, but also address legal and societal questions.
- Jonathan Sprinkle
- Electrical and Computer Engineering @ UArizona
- Model-driven software development (WP)
- Model-driven engineering (WP)
- Defense Advanced Research Projects Agency
- darpa (WP)
- DARPA Grand Challenge (WP)
- Artificial intelligence (WP)
- Google driverless car (WP)
- Electronic Control Unit (WP)
- Lidar (WP)
- Gyro (WP)
- Kalman filter (WP)
- Machine learning (WP)
- Artificial neural network (WP)
- State diagram (WP)
- Autonomous cruise control system (WP)
- Lane departure warning system (WP)
- Episode on Systems Health Management
- Component-based software engineering (WP)
- CAN bus (WP)
- ROS – Robot Operating System
- Time-triggered architecture (WP)
- Time-triggered CAN
- Dependability (WP)
- JAUS (WP)
- Vanderbilt’s GME
- MATLAB – The Language of Technical Computing – MathWorks
- Matlab (WP)
- C (WP)
- C++ (WP)
- mbeddr.com | engineering the future of embedded software.
- Security (WP)
- Security through obscurity (WP)
- Onstar (WP)
- Google driverless car (WP)
- Press Release: Carnegie Mellon Creates Practical Self-Driving Car
- The Penn State Autonomous Surface Vehicle Project (link dead)
- Rio Tinto: Home
- riotinto (WP)
- flarm (WP)
- Intelligent transportation system (WP)
Pingback: Vorhersage Sonntag, 02.02.2014 | die Hörsuppe
Highly recommended is the “Progamming a Robotic Car, CS373” class at Udacity.com, taught by Sebastian Thrun. You will learn a lot (A* search, Kalman Filters etc.)
Thanks, this is a great suggestion!
Good introduction and you touched on a lot of interesting topics. The focus was clearly on the systems and society perspective and you nailed that.
I would love to see another episode focusing more on the CS and societal angle of Artificial Intelligence in general. E.g. an overview which kinds of algorithms are actually used for certain tasks and where their strengths and limitations lie. Just a coarse overview ofc.. I’d also love to hear an experts opinion on how far away we are from strong AI and what “Intelligence” even means. I’ve read Ray Kurzweil’s “How to create a mind” (http://en.wikipedia.org/wiki/How_to_Create_a_Mind) and he makes it sound like we most likely have all the parts and only need to figure out how to properly put them together and some other perspective on that would be interesting.
Thanks to everyone involved in ωt. This podcast is definitely the best out there.
Thank you nononono …. :) Wrt. to AI, did you hear the interview with Ray Mooney? Some of the topics you mentioned are in there:
I think the civilian “killer app” will be autonomous taxis. It is expensive to pay drivers 12 or 24 hours per day to keep a “manual” car on the road – a large part of which is spent waiting in a taxi rank. And it’s dangerous in some areas.
If you leave out the driver, you can pay for the fancy electronics – and you can fit an extra passenger.
The big question: Will the public trust them?
Good point. Though I am not sure that taxi drivers are paid for the waiting time :-)
I’ve known some people who work in this area and am excited for the future. I do wonder though if we’ve passed the point of ‘does it work’ and onto finer details such as can we detect when we can poll the sensors less frequently or even shut some off to conserve operating efficiency. Further, assuming we’re moving towards complete autonomy, how would sensor failure or malfunction be handled?
The security systems you’re discussing for cars sound similar to methods people are trying to put into RFID. Apparently a ton of papers have been published on the subject since I first read about it years ago, but hopefully the auto security measures will actually be implemented since the technology hasn’t become mainstream yet.
On a final note, a lot of people care about the aesthetics of their car. If car manufactures start integrating these sensors, do the limitations of the technology require they be so obtrusive? As in, must they be tacked onto the roof and grill? From a quick search, it appears as though Nissan has seamlessly integrated a lot of these sensors without them sticking out like a rear-view mirror, but I don’t know if it’s as robust as Google’s.
I think we will get used to the changing aesthetics. We also got to like the new streamlined, windtunnel-optimized shape, didn’t we?
Please also note that driverless cars currently only work in sunny California. Bad weather conditions? Rain, snow, fog? No lane markers? No way to use GPS for now, not accurate enough.
Weather is a good point. Didn’t think to ask about that….
Another great episode, but still left me wanting more. I’m very curious to see examples of what the DSL looks like — i.e. what level is it written at? And how many lines of DSL are there to control the car? Great stuff about event-triggered vs. time-triggered, but would be very interesting to know examples of the events. Also, I think someone said that in a car, if you get into a situation where you don’t know what to do, you can always stop, unlike an airplane. But, if you are zooming down a freeway, seems like stopping could be dangerous, too.
I will forward this to Jonathan for the DSL examples.
But yes, you are right, stopping is certainly not always simple or feasible :-)
Pingback: Omega Tau Podcast Interview with FLARM Developer | Soaring Cafe
Pingback: Fast and Reckless Infographic
Great stuff. Thx. I mean I´ve been working in the automotive world for a while. I thought that I am quite experienced. But Jonathan gave me some interesting eye-openers. Thank you! It is good to get fresh input from time to time.
But honestly I am thinking that we should think globally about the self driving cars. But in the detailed requirements we have to think locally. I mean I am german but I´ve been living in Spain since 2010. And I can see definitely a difference in the driving behavior and rules between Germany and Spain. Anyway I am looking forward to see self-driving cars. and for the next episode from OmegaTau.
I see your point with the local details. But I am not sure we want to program autonomous cars to drive like Italians (at least not those in Naples I experienced during my vacation two years ago) :-)