MIT Tesla/Vehicle Technology Research

[MITAgeLab]

MIT AgeLab, a part of the Center for Transportation and Logistics at MIT, is beginning a new research initiative that analyzes advanced vehicle technology use in real-world scenarios. In particular, we are interested in which technologies people use/do not use, how people learn and interact with each technology, and which technologies promote safe driving.

For this project, we would ideally find drivers of recently purchased vehicles and install our removable, small form-factor, non-intrusive data-logging equipment in the vehicles. We would love to get Tesla drivers on board, since Tesla has the most cutting edge version of the advanced technologies we are interested in studying. This equipment would record things like audio/video from the cab, acceleration data, gyroscope data, GPS position, and a wide array of vehicle telemetry (velocity, wheel position, state of alarms and active safety systems, etc.). Privacy is very important to us. All data would be stored securely at MIT, and any results would be published without personally identifiable information. We would need access to your car to install and remove the equipment, and may need to do data retrieval/equipment maintenance once every other month or so.

While we are not yet installing this equipment, we did want to gauge general interest and concerns with our study. Is this a project that you’d be interested in hearing more about? What concerns would you have? If you were to participate, what would you consider to be fair compensation?

I look forward to hearing any input you have. I will be following this thread, but you are also welcome to contact me directly at [email protected]. If you have interest in the technical aspects of this study, especially related to the underlying machine learning and computer vision algorithms, feel free to reach out to my colleague Lex at [email protected] - he would love to hear from you. Thanks so much for reading!

Hillary Abraham
Research Associate, MIT AgeLab

 

[autodidaddict]

You had me until you mentioned audio/video from the cab. I'm all for research and helping fellow enthusiasts, and I generally allow far more anonymous data gathering about my movements than most of my counterparts in IT, but no way I'm allowing remote storage of video and audio recorded from inside my vehicle. This is a deal breaker for me and I suspect I would not be alone in this.

 

[MITAgeLab]

Hi autodidaddict,

Thanks so much for your comments. We completely understand that the recordings might be a deal breaker for some or many individuals. Part of the goal for this post is to find out how many individuals would refrain from participating in the study for specific reasons, such as discomfort from being recorded, so this is very valuable feedback.

If you do not mind, I’m curious to know if your issue is with the information we are collection (audio/video in this case) or with the fact that we are storing it remotely? That is to say, would you be more comfortable with us collecting this type of information if you had more information about the type of security we use for housing this data, or are you uncomfortable with it under any circumstances? We are still in the planning stages of this study, and really appreciate the advice!

 

[CTShore]

You had me until you mentioned audio/video from the cab. I'm all for research and helping fellow enthusiasts, and I generally allow far more anonymous data gathering about my movements than most of my counterparts in IT, but no way I'm allowing remote storage of video and audio recorded from inside my vehicle. This is a deal breaker for me and I suspect I would not be alone in this.

I agree. For example, recording phone calls made/received in car? No.

 

[notice]

I would be concerned with the installation of your monitoring equipment and how it might void the Tesla warranty and/or cause problems. For example, in these forums, there is a case of someone having a professional dash cam installation that caused major reliability problems in the car due to a bad splice in a wiring harness not covered under Tesla warranty to diagnose and repair. Is your installation authorized by Tesla Motors? Do you guarantee any expense to fix any problems at your expense that your installation may cause? Are their sufficient cash reserves in your budget and/or escrow to cover such expenses which could reach many thousands of dollars per vehicle? And I echo the concerns of @autodidaddict - video and audio monitoring seems above and beyond any basic telemetry you may be acquiring for your research and represents a major invasion of privacy. Personal and private conversations, confidential and sensitive information about business conducted in the car, or personal habits which your researchers will have unfettered access to is disconcerting. Do they all sign my confidentiality agreements? What is your data retention policy? How can you guarantee the security and privacy of the data stored (in a world where major hacks make national news every month)? Will other researchers have access to this raw data now and in the future? Would you be compelled to turn over this data in whole to third parties if subpoenaed? Don't get me wrong, the field of research is interesting and important, and I have many MIT connections. But tapping into the generally undocumented systems of my very expensive car and subjecting me to constant unfettered surveillance is a sacrifice I am not prepared to make in the name of science.

 

[sbanks]

My sole concern would be indemnification against any vehicle damage due to installation/removal of your equipment. I don't have any issue recording audio/visual inside or outside my car; I'm really not interesting enough to worry about anything you may record.

 

[wk057]

Yeah, the AV recording is definitely a show stopper for me, personally. I don't see how that data could be of value, especially audio recordings from the cabin. I have conversations in my vehicle regularly with passengers and via the phone that I would consider private or otherwise confidential that I simply could not allow a third party to have access to. There are potential legal implications here as well, such as laws surrounding phone recordings and third party notification (I keep the audio recording on my dash cam disabled for this reason). In most states people would be legally required to let other people know that they were being recorded by a third party. I think you'll definitely have a much easier time with this if you forgo the AV recording aspect.

Further, with the Model S in particular, to get data such as wheel position and accelerator position you'd have to utilize one of the diagnostic connections stashed under the center screen (above the pocket there) to monitor and decode high speed CAN activity. Either that, or you would need Tesla's full blessing and support for using the diagnostic ethernet connection (locked out via public/private key encryption, IIRC). Data like that isn't available elsewhere, unless you were installing your own sensors.

Tesla seems to be very against third party modifications that access the CAN buses, such as 3rd party Mobileye installs on the non-autopilot models. They're also mostly against 3rd party modifications of any kind, it seems, where things tap into constant power. So, I think any agreement with Tesla owners would definitely need to include indemnification against damage from your install/removal as well as indemnification against potential issues with Tesla service should anything related to this equipment cause concerns with Tesla down the road. Our warranties are pretty important. A waiver probably won't cut it here.

Keep us informed.

 

[MITAgeLab]

Thank you all for your thoughtful responses! The feedback you’ve given is invaluable to our success. It seems like the factors that would most likely prohibit participation in an experiment like this one are privacy concerns related to audio recordings and data storage and the possibility of damage to your vehicle. All are extremely valid concerns. Our team has revisited these topics, and we are hoping you may have more thoughts for us.

First, we discussed removing the audio/video component from the study entirely. Privacy is very important to all of us, and we certainly understand wariness from being recorded in your own private vehicle. After discussion, we still believe there are a number of things to be gained from A/V recordings. We are particularly interested in your vocal interactions with technologies in the car and the situations in which you are using the technologies, which are most easily recorded through A/V equipment. For example, do you need to repeat yourself when interacting with voice systems? How does that impact your attention to the road and traffic? How are you responding to auditory safety alerts? In what traffic and weather conditions are you most frequently using different types of advanced technologies, and how do these conditions affect ease of use? While these questions are important, we do understand the concern for protecting confidential conversations and phone calls. We have added to our protocol the option to withdraw any or all audio recordings from our dataset at the end of the study. Would knowing that you could remove confidential conversations from the dataset affect your willingness to participate?

Second, we examined our policies for data storage and access to said data. MIT has very strict policies for working with human subjects and collecting data, and we constantly work to ensure privacy and confidentiality are maintained. These policies are outlined in depth at couhes.mit.edu , but I will go over the top-level relevant information here. All staff that work with any human subjects data are required to be authorized on their project and go through extensive training every three years. MIT is obligated to share data collected with other third parties as required by law. For all of our studies, most of the data (everything except for A/V recordings) is stripped of any personally identifiable information and only published in aggregate form. Data that does contain personally identifiable information is encrypted and stored within 12 hours of collection, and in the interim must be kept in a locked bag or carrying case. Since the data collected will have the potential to answer research questions far beyond the initial scope of this work, we do plan on maintaining the data in perpetuity. However, as noted above you would be welcome to withdraw any audio from our dataset as you see fit.

Third, we reviewed the possibility of damaging your vehicle. We will not be modifying any aspect of the car in a permanent way. We will be attaching a number of cameras with double-sided tape on surfaces where such tape does not leave any marks. We will hide the cords to the logger to the best of our ability such that the visual appearance of your Tesla is still great (as it always is). We will connect to the diagnostic port of the car for power and to get CAN information, but this will require no modifications to the car and will not cause any damage.

Finally, we took these risks into consideration and came up with a payment structure for participation in the study. As mentioned earlier, we would need access to the vehicle to install and remove the equipment, and may need to retrieve data or service the equipment once every other month or so. We expect participants to earn approximately $1000 for every year of participation in this study.

With these amendments, do you have further comments? Thank you all once again for your attention and advice as we move forward with this project.

 

[trils0n]

Might be important to note that the OBD2 diagnostic port in the Model S provides no data (it does have 12v power, however). It is only there because it is required by law, but it isn't connected to anything, so you can't get data from it.

 

[MITAgeLab]

Thank you - that is good to note! We have a Model S on-site that we are using to design the logger, and we are able to pull CAN info from the diagnostic port.

 

[wk057]

The diagnostic port does have some CAN connected. It has the body CAN bus, and a bus that is just for the OBD port from the MCU. It doesn't have access to the power train CAN and the other buses. For those you need the diagnostic connector above the little cubby under the 17".

 

[David29]

In principle, I'd be interested in supporting the research. But I echo others' comments on the audio/video recording (particularly recording of people and conversations within the cabin, not so much video of the road, etc.). I seriously doubt my spouse or any other passengers would want to have recordings made, and it might alter who was willing to ride in my car. I am not sure I could allay everyone's fears on the spur of the moment when offering them a ride, and asking them to sign a form (if required) would be too intrusive, I think.
one had installed a wire that ran along the lip of the sunroof opening under the headliner. So I have even refrained from installed the EVannex after-market coat hooks for fear of violating the warranty. (My concern is amplified because this is a leased car, and the lease has its own restrictions about violating warranty.)
But monitoring the car itself seems OK.

 

[AB4EJ]

The Virginia Tech Transportation Research Institute has done quite a bit of research in this area, also including in-cab recording of audio & video. They have also published some papers about this, I believe under the aegis of the Center for Automated Vehicle Systems, which studies human factors as related to vehicle automation capabilities.

 

[Bluespark]

I don't quite get what would be gained by audio/video input. Actually I don't see where wheel position or any vehicle parameters would help you learn how we interact with technology/our cars. We get in and drive. It has a brake pedal and an accelerator, like any automatic transmission car. Where we drive (GPS tracking) has no bearing on what we are driving. No offense, but I think you are looking too deep into this. The really advanced technology that our cars have deals with the batteries, the battery management system and the drive system. The full extent of our usage/involvement of this advanced technology is the use of the accelerator with our right foot.

Sure the car has a 17" I-Pad for the center screen, but I am sure you haven't done research on I-Pad users....

 

[rideralong]

Driver distraction is the next thing, especially in cars that can drive themselves - to some extent. Now only to find out what distracts us while driving.....

I agree with most others about video and audio recording in the car. Although, if we're given the option to remove given segments of the recordings that would make me feel at ease much more.

Looking forward to seeing where this goes!

 

[MITAgeLab]

Hello again, TMC! We wanted to sincerely thank you for your feedback in the fall, and reach out with a brief update on our study progress so far.

After modifying the study based on your suggestions, we went live in late January and have since instrumented and are successfully collecting data from 4 Teslas. Based on our progress so far, we expect to have 12 Teslas instrumented by the end of this year! Our lab and this project were recently featured on WBUR here: http://www.wbur.org/2016/04/29/traffic-future-driverless-cars.

The work we’re doing is, among many other things, helping us improve driver gaze classification. Our team made a short video demonstrating our methods here:

We invite any of you who might be interested in participating in the study to contact us at [email protected]. More information about the study can be found at: http://tesla-study.mit.edu

 

[jmanning]

I am one of the people involved in the study and have had no problem doing it. It hasn't impacted my life at all. The wbur piece I happily took part in. If you have any questions feel free to PM me. I encourage anybody who has autopilot to join in this fantastic research !!

 

[rz-1]

I'm very interested in this research and have already contacted the MIT team. I have no problem with collecting the video, as that is essential for studying the driver distraction and the related topics (glad MIT is doing this research)! I shiver every time I see the crazy videos of the way people engage the Autopilot and wish the car could sense more than just my grip on the steering...Audio is a concern, especially if I'm having a work related call, etc. (which now seems to be optional). But by far, my concern is regarding the potential issues with the warranty.

The MIT team provided me with an honest response (as also indicated here), regarding the work they do on the car and the fact that no one has had any issues so far. I called both service centers in Boston area. The Dedham center confirmed they have seen 4 cars instrumented for this research and there has been no issues. But they also said that "if" I encounter an electrical issue with the car in future, when they see the aftermarket equipment, they will examine if that work may have caused the electrical problem. It is only then that I may run into some issues with the warranty.

As MIT tries to expand this work, would be nice to come up with some agreement with Tesla (e.g. pick an approved installer) and completely eliminate this worry.

For now, I need to do some more research on this and see where I land! I'm leaning towards going ahead with it...

 

[MITAgeLab]

Yes, audio recording is optional. Drivers are also permitted to delete any portions of the data they do not want included in the dataset by emailing research staff a date and time range for removal. So for example, if one was generally OK with audio recordings but concerned about the occasional phone call, those segments could be deleted. If one didn't want to worry about emailing us after a phone conversation, forgoing audio recordings entirely is fine as well.

I'm very interested in this research and have already contacted the MIT team. I have no problem with collecting the video, as that is essential for studying the driver distraction and the related topics (glad MIT is doing this research)! I shiver every time I see the crazy videos of the way people engage the Autopilot and wish the car could sense more than just my grip on the steering...Audio is a concern, especially if I'm having a work related call, etc. (which now seems to be optional). But by far, my concern is regarding the potential issues with the warranty.

The MIT team provided me with an honest response (as also indicated here), regarding the work they do on the car and the fact that no one has had any issues so far. I called both service centers in Boston area. The Dedham center confirmed they have seen 4 cars instrumented for this research and there has been no issues. But they also said that "if" I encounter an electrical issue with the car in future, when they see the aftermarket equipment, they will examine if that work may have caused the electrical problem. It is only then that I may run into some issues with the warranty.

As MIT tries to expand this work, would be nice to come up with some agreement with Tesla (e.g. pick an approved installer) and completely eliminate this worry.

For now, I need to do some more research on this and see where I land! I'm leaning towards going ahead with it...

Thank you for the feedback! We would love to pick an approved installer when moving forward with this work, particularly when considering a nationwide expansion. It is definitely something we are considering and welcome suggestions from the TMC community or others.

 

[JCL123]

I just contacted them since I got my Model S recently. I may be up for it.

-JCL