In the first in a series of seven articles sponsored by Chevron, reporter Jacob Stoller looks at the role artificial intelligence could play in repair bays of the future.
By Jacob Stoller
Many shop owners view anything high tech with trepidation… and with good reason.
The digitization of today’s vehicles has moved diagnosis outside the comfort zone of many veteran technicians. It requires continuous investment in proprietary training and tools. A shop can tie up a top technician for hours troubleshooting an electronic fault and barely break even on the effort.
Does it make sense to apply the highest of high technology to maintaining and fixing cars?
“Definitely,” says Alexander Wong, University of Waterloo professor, and Canada Research Chair in Artificial Intelligence and Medical Imaging. “Right now, AI is being used more by the manufacturers. But using it in independent service shops could actually work out quite well.”
An AI tool, for example, could substantially reduce the time technicians spend troubleshooting. It could remove knowledge barriers, making it possible for many more technicians to diagnose complex problems. It could improve efficiency of troubleshooting by prioritizing steps and tests. It could give shops new capabilities, such as being able to anticipate mechanical or electronic failures and warn customers to take proactive steps.
Such tools are not yet within the means of independent shops, but the technology is definitely moving in that direction.
Kitchener,Ont.-based Acerta Analytics Solutions Inc., an AI start-up spun out of research activities at University of Waterloo’s research and development labs, is currently helping automotive OEMs apply AI to connect the dots between failures and various conditions such as temperature or pressure.
“We’re applying AI and machine learning to the data being generated from the various sensors and computers within a vehicle,” says Greta Cutulenco, Acerta CEO and co-founder.
What the OEMs have realized, Cutulenco explains, is that their engineers don’t have all the answers to failure-related issues when they put their cars on the road. They have to acquire this intelligence from the field.
To achieve this, they’re collaborating with their dealer shops to gather and centralize the data. Then they apply Acerta’s AI technology to establish the applicable patterns and correlations. The information is used to improve dealer shop diagnostics, assist with predictive maintenance, and improve vehicle quality.
Another AI company involved in the field is Mississauga-based CaseBank Technologies Inc., a Division of ATP, which has its roots in aviation but has recently expanded into the automotive space, and now supports a global automotive manufacturer.
“They’re particularly interested in gathering that field experience, because that’s something that they don’t otherwise get,” says CaseBank chief technology officer Mark Langley. “They get it at the dealership level, but not back to home base. So this gives them some visibility into the field, and allows them to influence their re-engineering efforts as well.”
CaseBank’s SpotLight product employs a branch of AI called case-based reasoning. When fed details of a current scenario, SpotLight compiles a list of similar scenarios from the existing database, ranked by similarity. SpotLight then recommends a test sequence to shorten that list and identify the best fix, using factors such as the track record of a particular fix in similar scenarios, the cost and time required to run particular tests, and the information that those tests provide.
“That’s the strength of AI. You don’t need to know exactly how each model of, say the transmission or the car, works. You just learn its typical behaviour, and then you can find similar patterns of failure across various types of vehicle models.”
“The real value comes in how you decide to order the test sequence,” says Langley. “You have to balance increasing certainty through further testing against the cost and time of doing so, and the cost and time of making the wrong decision.”
Both Cutulenco and Langley believe their technology is well suited for independent shops. The key here is that AI essentially learns on its own from the data, as opposed to having to be pre-programmed by engineers.
Audi’s Remote Telepresence (ART) a remote robot on wheels helps an automotive technician.
AI that has been developed and commissioned by a non-OEM player would learn in essentially the same way as one commissioned by an OEM. This means that under the right circumstances, an AI could collect data and acquire predictive maintenance and problem diagnosis intelligence using vehicle data independently of OEMs.
Independents, unlike OEMs, could aggregate data across vendor lines, and this could be an advantage. “I would say that across the different product lines, failures do arise that are very similar and have very similar patterns,” says Cutulenco. “That’s the strength of AI. You don’t need to know exactly how each model of, say the transmission or the car, works. You just learn its typical behaviour, and then you can find similar patterns of failure across various types of vehicle models.”
Getting the Data
There is, however, a major catch. OEMs keep tight control over their data.
OEM scanners, for example, don’t allow their data to be exported to external devices or databases, so an independent database would be limited to ODB2 codes and other industry-standard information available on the CAN bus. And although it’s technically possible to scan an OEM troubleshooting manual into a database, repurposing it in an AI might violate copyright laws.
In the worst case, an AI could start with currently available information, and then build up a knowledge base on a case-by-case basis. This is already being done. CaseBank has built up a database independently of OEMs. The data comes from manually written reports from airline pilots and ground maintenance personnel, thus avoiding the use of any proprietary information.
An AI tool could substantially reduce the time technicians spend troubleshooting. It could remove knowledge barriers, making it possible for many more technicians to diagnose complex problems.
A consortium of shops could probably undertake a project like this, says Wong, and things are likely to get easier. There are reportedly companies that are trying to crack the OEM codes, and it’s likely a matter of time before more data becomes available.
“Over the next five years or so,” says Cutulenco, “as data will become more consistent and of higher quality, they’ll definitely be able to start applying these types of technology.”
Another possibility is that aggregators will enter the marketplace and make AI capabilities available to independents.
For example, tool and services company Snap-on has engaged California-based AI provider Predii to extract diagnostic intelligence from their database that contains literally billions of service events.
Knowledge is Power
AI is a double-edged sword. Under current circumstances, it appears AI could become one more weapon in the OEM knowledge advantage.
On the other hand, the technology is developing very rapidly, and new players are continually emerging. AI technology is already well-suited for helping independents out of the diagnostics conundrum, and it’s likely that new players will emerge to serve the thousands of auto shops that could benefit. And if enough independents can find ways work together and share their data, technology developed right here in Canada is already available to turn that data into a competitive advantage.
Jacob Stoller is a freelance writer living in Toronto. He specializes in technology and Lean management. You can reach him at firstname.lastname@example.org