[Ravich]

I am seeking inputs regarding the value and possible interest in an open source Engine Analytics project amongst team-bhp members.

The value proposition for the project :
1. There are a significant number of members who would like to monitor and compare their engine performance and driving styles with those of fellow members. This means they would agree to upload anonymised engine, GPS data to a cloud based storage and analysis engine.

2. They would find value in a Black Box like capability to be used in case of road accidents / incidents , catastrophic engine or mechanical failures
3. They would be willing to try out a Raspberry or Beaglebone based OBD II data collector , GPS (optional)antenna, dashboard , Dashboard cameras (front and rear) and dashboard display .
3. They would be willing to upload via 3G , data to (possibly AWS based) analysis servers
4. They will find value in comparative analysis and insights on engine performance (including condition monitoring where possible) provided on platform independent browsers ( chrome, Firefox, Safari)

5. They would be benefit from incident analysis provided on demand
6. The analysis engine would notify them of engine degradation or other anomalies detected using fleet wide data, specific to their models

The project would be developed around Open source Beaglebone Black or Raspberry Pi and low cost associated OBD II ELM chipsets/adapters available and the Python ecosystem. (PyOBD etc..) The analysis engine would run on AWS or equivalent and would leverage open source analytics tools.

Clearly this is not a new idea but I am not aware of any powerful analytics for cars that provide insights derived from fleet level data ( atleast not open source).

Please let me know of your interest, expertise and willingness to participate. I am currently based in Bangalore. All source ( HW/SW) would be open .
Regards

[Foxbat]

Sounds interesting but what type of help are you looking for? I have about 6 years experience in Data Analytics in CPG, Retail and Banking.

Unfortunately none on automobile analytics

[DerAlte]

Why "Open Source" or "Cloud Storage" - is that the driving factor? Don't want to sound negative or rain on your party, but putting things together like this one can only get vicarious pleasure in an amateur college project design effort, like getting Raspberry Pi to do something. Such keywords are meaningless and distracting unless the primary goals are significant and practically achievable. In a design exercise, one can't treat data like an abstract collection - even though analytics will deal with it later to produce abstract human-understandable results.

The devil is in the details. Objectives cannot be vague - "They will find value in ..." is usually a speculative proposal made by Marketing, who have only a peripheral relationship with the actual equipment / vehicles. One can see this in commercial software products, where 90% of features are fluff that no one uses.

I have been with the subject for a little more than 20 years (the subject Vehicle Informatics itself is older than that), and have a fair idea having been a participant. Open Source, cheap hardware, data capture / communications / analytics etc. are hardly significant issues. The main issue is sanguine attribution of cause v/s effect. Otherwise, one lands up having a ton of data - and still misses the forest for the trees.

Fleet level analytics is a completely different subject altogether. That deals with parameters that ultimately have an effect on costs: fuel / consumables (oils) / perishables (tyre) consumption. No fleet owner wishes to maintain vehicles for the design lifetime of the vehicles, since it is common knowledge that maintenance costs grow exponentially beyond a point of time. Rental car agencies in Europe and US typically sell cars off after 1 year. Maintaining them beyond this point has a very large organizational cost, with no commensurate returns. Freight operators retain vehicles longer, since the design life of trucks is much longer (due to simpler systems) than that of cars.

[NetfreakBombay]

Originally Posted by Ravich

The project would be developed around Open source Beaglebone Black or Raspberry Pi and low cost associated OBD II ELM chipsets/adapters available

Focusing on software (instead of hardware) might work for this project. Significant % of target audience would already have a smartphone, all they would need is reliable OBD II scanner that can interface with smpartphone.

With that, Cloud can become an option. User can also choose to store data on phone.

[Ravich]

Originally Posted by DerAlte

Why "Open Source" or "Cloud Storage" - is that the driving factor? ....

Such keywords are meaningless and distracting unless the primary goals are significant and practically achievable.
.....

Objectives cannot be vague - "They will find value in ..." is usually a speculative proposal

.....
Open Source, cheap hardware, data capture / communications / analytics etc. are hardly significant issues. The main issue is sanguine attribution of cause v/s effect. Otherwise, one lands up having a ton of data - and still misses the forest for the trees.
.....
Fleet level analytics is a completely different subject altogether. That deals with parameters that ultimately have an effect on costs: fuel / consumables (oils) / perishables (tyre) consumption.

Thanks ! Appreciate the comments. The post itself was to validate some of these leap of faith assumptions.

Highlighted the Open Source to convey the non-commercial intent ( the spirit of Team-bhp) behind the initiative and to emphasize the availability of powerful ecosystems available to try this out . Cloud was to emphasize the feasibility of scaling to accommodate large number of participants and vehicle data at manageable costs.

Primary goal or value proposition to be validated are :
As an enthusiast I would like to compare my cars engine performance over time with that of other members having similar/same models ( what I call the fleet data in a narrow sense)
The engine and other data can provide early warning of potential problems ( prognostics) which would be of value to enthusiasts if the cost of of such prognostics is not high . Hence the focus on open source HW/SW .

The point about cause/effect and tons of data being useless is very valid. I would submit open source analytics now makes possible detection of anamolies not necessarily from physics based first principles but from data obtained from a number of 'similar' vehicles . That is the reason why a large number of participants would be useful - which supports the low cost .

As I mentioned, I used fleet in the sense of data from a number of vehicles of similar make rather than operating a fleet .

All that said , I certainly would appreciate if members with rich experience can participate and your vast experience would certainly be a great help. I think it would be a good cause and importantly a fun one .

But the intent was really to validate some of these value propositions.

[Ravich]

Originally Posted by NetfreakBombay

Focusing on software (instead of hardware) might work for this project. Significant % of target audience would already have a smartphone, all they would need is reliable OBD II scanner that can interface with smpartphone.

With that, Cloud can become an option. User can also choose to store data on phone.

Thanks for the feedback !
The value story is the comparison of your engine/vehicle performance with similar ones and anamoly detection from 'fleet' data. Hence the need yo upload data to a common storage. Cloud for scalability and accessibility . No HW will be developed as such .

[Ravich]

Originally Posted by Foxbat

Sounds interesting but what type of help are you looking for? I have about 6 years experience in Data Analytics in CPG, Retail and Banking.

Unfortunately none on automobile analytics

Thanks for the offer !. The analytics as I mentioned is not necessarily from first principles. So any analytics expertise is only going to strengthen this.

[amit_mechengg]

Nice initiative. On the similar lines there is one more company manufacturing CAR IQ product. It collects all the vehicle data through OBD, connects with piers and other agencies via internet and shares data.

[DerAlte]

Originally Posted by Ravich

... As an enthusiast I would like to compare my cars engine performance over time with that of other members having similar/same models ...

Despite the fact that this varies so widely (due to differences in driving styles and conditions) that comparison is futile?

Comparing with 'similar' models? Well, well, well!

Originally Posted by Ravich

... The engine and other data can provide early warning of potential problems ( prognostics) ...

Sir, IMHO you should first read the different threads where people describe their experiences and problems with different cars. Readymade data, almost none of which is a harbinger of potential problems to others. It causes paranoia in some cases, like 'what if it happens to me in the middle of a long deserted highway?'

Originally Posted by Ravich

... but from data obtained from a number of 'similar' vehicles . That is the reason why a large number of participants would be useful - which supports the low cost . ...

Sir, how many things do you buy because the price / cost happens to be low? Sure, a lot of parents buy their children cheap sunglasses with pink / yellow / green / blue lenses - cheap novelty that amuses children.

From comparing similar data p-o-v, here is a rather amusing example. This should highlight the risk of abstracting data without sanguine attribution and rational correlation.

[Ravich]

Originally Posted by DerAlte

Despite the fact that this varies so widely (due to differences in driving styles and conditions) that comparison is futile?

Comparing with 'similar' models? Well, well, well!

Sir, IMHO you should first read the different threads where people describe their experiences and problems with different cars. Readymade data, almost none of which is a harbinger of potential problems to others. It causes paranoia in some cases, like 'what if it happens to me in the middle of a long deserted highway?'

Sir, how many things do you buy because the price / cost happens to be low? Sure, a lot of parents buy their children cheap sunglasses with pink / yellow / green / blue lenses - cheap novelty that amuses children.

From comparing similar data p-o-v, here is a rather amusing example. This should highlight the risk of abstracting data without sanguine attribution and rational correlation.

I would have liked a few more questions on what could be the data , the possible analytics and whether there are some examples where such data driven analytics have been applied with some degree of success, before the ill-disguised sarcasm.

But I do have very high regard for the Team-bhp members and the spirit in which it is run, to still expect some inputs on how this could be refined (or 'pivoted').

Leaving aside the possible analytics/prognostics , imagine the usefulness for the consumer if the aggregated data from a specific model from a statistically significant number of users reported an unusually high number of say fuel system leaks in the trouble codes logged by the OBD.

Whether this project is useful or not, logged (OBD and other) data will be the norm in the future and offer great possibilities. As a sensational example, I am sure many have read about the famous Elon Musk rebuttal of Top Gear's Tesla Roadster's test issues with the data.

We can find a thousand reasons why this may not be a good idea - but it is always interesting to find a few that tell us why this or how this could be a good idea. Sorry if it sounds cliched - I refuse to be cynical.

[Ravich]

Originally Posted by amit_mechengg

Nice initiative. On the similar lines there is one more company manufacturing CAR IQ product. It collects all the vehicle data through OBD, connects with piers and other agencies via internet and shares data.

Excellent tip Amit - I just looked up - I am certainly going to be rooting for them ! I intend to pre-order one .

[DerAlte]

Originally Posted by Ravich

... ill-disguised sarcasm. ... I refuse to be cynical.

@Ravich, when someone else is genuinely amused and expresses to the contrary to what you think, it is not 'sarcasm' ('there are many other sadnesses in life' - as the saying goes in Hindi - for others to indulge). It is only your refusal to mentally accept, that what you are proposing is not on the right track, which makes you term it that way!

And you should not be 'cynical' at all - being 'pragmatic' is a better state to be in, by all means. Just make sure you are in touch with the facts and realities of this subject. Having nebulous objectives and wishes don't get one anywhere. All are searching, no one has all the answers yet, everyone is speculating from different p-o-v. If something happens to be working from one's p-o-v (without corrupting it), one is lucky.

[maverickgoan]

Very Interesting thread. have you made any progress since the last post on this thread ?
Had the same thoughts for a long time & have a pet project to capture data to record & analyze my driving style & how i can improve myself. ( i have about 5-6 analysis points )
e.g. do i accelerate just before braking ?

I have skill set with micro controllers (raspberry, Arduino), ODB ports & general electrical wiring but little experience with software coding. All the efforts I put in the past to learn how the OBD protocol works got reset when I changed my car last year.

[Rehaan]

I had come across this video on Data Acquisition & Performance Tuning (for racing) years back.

It might require a Silverlight plugin on your browser to play (try firefox).

1 hour long, but i remember it being really good!

http://mediasite.engr.utexas.edu/UTM...7814017477011d

[maverickgoan]

here is something i just came across. https://create.arduino.cc/projecthub...g___&offset=18

@ravish, feel free to PM me if you want to work this idea. i have some free time now a days.