Cycling isn’t an expensive hobby outright. But as you start getting serious, a cyclist can spend a frankly exorbitant amount of money measuring what their legs are doing. In a more professional setting, things like instrumented pedals, bike computers, heart-rate sensors, and a lot more start becoming important. All of this just helps squeeze out more information out of every ride.
Now, researchers at La Trobe University (Via TechXplore) are exploring whether one piece of hardware most people already own could eventually handle a surprisingly sophisticated part of that job. This isn’t some specialized hardware either; it’s a small gadget that fits in your pocket–and yes, it’s your phone. Researchers at the Holsworth Biomedical Research Centre are developing AI models capable of estimating the forces a cyclist applies to the pedals using motion captured on video. A rider can simply record themselves with a smartphone and receive biomechanical information that currently requires much more specialized equipment.
AI is watching every pedal stroke
The system uses deep learning to connect what it sees in a cyclist’s movement with the actual forces measured at the pedals. For this research, they trained the model using synchronized laboratory recordings, which include one dataset of video recordings of the rider in motion, while another measures the real forces being applied during each pedal stroke. The AI then learns the relationship between the two and attempts to predict those forces from movement alone.
And we’re already seeing some promising results. An earlier study involving La Trobe researcher Rodrigo Bini used recurrent neural networks to predict three-dimensional pedal forces as well as forces and power at the lower-limb joints. This report showed correlations between measured and predicted variables ranging from 0.79 to 0.96, though accuracy varied depending on the force or joint measurement being estimated. So the new project could make this sort of analysis easier to scale outside the lab.
Your phone isn’t replacing a power meter tomorrow
This research remains at an early stage. La Trobe isn’t dropping an app anytime soon. Bini estimates that within five to 10 years, smartphone apps could potentially let almost anyone record their movement and estimate cycling forces. The team is now strengthening and validating its dataset, with preliminary findings expected to be presented at the International Society of Biomechanics Conference next year.
We have tested smart pedals in the past, like the Garmin Vector 3, that are capable of telling you where and when you generate force during a pedal stroke. However, this kind of granular data has traditionally required dedicated sensors and serious money. Turning a phone camera into even an approximation of that toolkit could put genuinely useful cycling biomechanics within reach of far more riders. With AI being used as fitness coaches as well, we are soon seeing tech being used to fine-tune more physical aspects of our lives.