COMMENT: Like any new worker, AI will need time and practice

It seems likely that automakers will need AI-enabled manufacturing techniques to build tomorrow’s cars, but they should expect some teething problems. By Xavier Boucherat

In the car factory of the future, a worker from the final assembly is being watched attentively by a heavy load-handling cobot, outfitted with a range of complex sensors. The worker demonstrates a complex assembly task. Using computer vision, the robot analyses what it sees, and creates programming which will allow it to perform the task itself. What’s more, it does not simply learn it by rote, but recognises the general patterns and laws around the procedure such that it can perform whatever slight variations in movement arc or otherwise are required to complete the task.

This is just one potential scenario being dreamt up by those who see a place for artificial intelligence (AI) in the future of vehicle manufacturing.

One thing is certain: building cars will only become more difficult as time goes on. Not only is the vehicle technology involved more sophisticated and intricate than ever, but the continued merging of multiple models onto single lines makes mistakes more likely

Whilst much remains uncertain about what form it might take, it does seem likely that deep-learning, advanced image recognition and other AI-related technologies will have a role. That’s because one thing is certain: building cars will only become more difficult as time goes on. Not only is the vehicle technology involved more sophisticated and intricate than ever, but customisation rates and the merging of multiple models onto single lines makes mistakes more likely. Robots may fail to correctly recognise a model and perform the wrong procedure, or fail to reset properly. Some intelligence will be needed to help make sense of the seemingly endless configurations coming out of factories.

All of this could arrive in tandem with the established industry megatrends: the rise of electrified vehicles with increasingly intelligent advanced driver assistance systems, eventually enabling autonomous driving, could prompt automakers to overhaul lines to maximise efficiencies. Then might be a good time to introduce new technologies, such as AI-assisted manufacturing.

AI may already be having an impact on our daily lives, a potent example being the rise of voice recognition systems such as Siri and Alexa, but those who regularly use these will know full well that they’re far from perfect

However, it’s important to remember that AI does not work like a standard algorithm, in that the way it works changes over time. Its results can be, by definition, unpredictable. Deep-learning technologies require colossal amounts of data for reference points and recognition. AI may already be having an impact on our daily lives, a potent example being the rise of voice recognition systems such as Siri and Alexa, but those who regularly use these will know full well that they’re far from perfect. Much higher levels of maturity and robustness will be required for a role in automotive manufacturing, where quality is critical for both safety and sales. In short, those automakers which do adapt AI can expect some teething problems.

The role of AI in the factory of the future is explored in depth in Automotive World’s latest special report, ‘Artificial intelligence and the future of vehicle manufacturing’, featuring analysis and insight from a range of industry stakeholders.

Close
Close