Could you harness the power of AI to enhance your life on two wheels?
Artificial Intelligence (AI) can rapidly crunch vast amounts of data to identify patterns and trends, and it’s creeping into every area of our lives, not least cycling.
Cyclists can harness this emerging area of technology to supercharge their performance, enjoy a more comfortable ride, or simply stay safe – and you may already be doing so without realising.
Here are five key ways in which AI is being integrated into cycling to enhance life on the bike.
Diving into data
AI enables fast analysis of performance data, such as speed, heart rate and blood oxygen levels, gathered from bike computers as well as wearables such as heart rate monitors and cycling watches.
This data is used by some of the best cycling apps to generate personalised insights to help cyclists improve their performance.
“It’s easy to assume that only pro racers are using machine learning technologies, but there are a growing number of consumer products using them to personalise training and extract new insights from the data you collect,” says Chris Ruddock, a wearable technology specialist.
“More are also incorporating AI interfaces that make data simpler and more engaging than ever.”
Ruddock cites TrainerRoad’s Adaptive Training platform as an example. “This recommends session plans based on how you complete previous workouts. This compensation is based on how you respond to each effort, theoretically removing the need to do Functional Threshold Power (FTP) ramp tests mid-block to recalibrate the intensity of sessions.”
The wearable WHOOP tracker, which counts Wout van Aert among its users, is another example.
WHOOP has introduced a Coach feature powered by OpenAI (of ChatGPT fame), which generates individual responses to users’ health and fitness questions based on their unique biometric data. For example, ‘Can you build me a 100-mile cycle training plan?’ Apparently, it can.
Optimising your calories
To minimise bonking, Team Jumbo-Visma (now Visma-Lease a Bike) has designed an AI model trained to predict the caloric needs of its riders.
Martijn Redegeld, the team’s performance nutritionist, explains that, in preparation for the 2023 Vuelta, the team entered core data (for example, power and GPX files), while the model sourced other input itself, such as weather conditions.
“With all this relevant input, the model starts to do its magic, and within minutes, we have the daily caloric needs of each rider accurately predicted for 21 stages,” states Redegeld.
“These predictions form an important starting point for the personalised nutrition plans generated in the team’s Athlete’s FoodCoach software, used by riders, nutritionist and chefs during a race.”
But it’s not only the elite who can fine-tune their sustenance. Another product, AI Endurance, calculates daily energy recommendations to maintain energy balance, and optimise nutrition, fuelling and recovery from training.
Increasing safety
Driver assistance systems are common in cars, but AI can also be used in the prevention of cycling accidents, analysing traffic data and identifying hazards.
For example, Copilot is a bike light and camera that monitors the road using AI to detect approaching vehicles.
Cyclists receive audible alerts based on the actions a vehicle is taking – such as whether it’s attempting to overtake or approach dangerously. It also syncs with a mobile app, enabling cyclists to see what’s happening behind them.
Clark Haynes, Founder of Velo.ai, the company behind Copilot, says the product has a key advantage over a continuous bike light. “The light on Copilot reacts to driver behaviour so, as a vehicle is approaching faster or coming closer to the cyclist, the light pattern will blink rapidly in response, increase intensity and change colour to draw as much attention as possible to the cyclist.”
The rear-facing camera records the ride, automatically pulling video clips of close calls or collisions. “It’s a huge advantage to immediately have access to a library of notable clips from your ride that downloads to your phone,” says Haynes.
Getting comfortable
Several bike-fit apps, such as MyVeloFit and Bike Fast Fit EZ, are using AI to crunch the necessary data. It’s a more convenient (and perhaps cost-effective) solution than visiting a bike fit specialist, but you’ll need a stationary trainer or rollers, as well as basic mechanical skills.
MyVelo uses two aspects of AI – Computer Vision and Pose Estimation – to analyse your position from mobile video footage.
“We specifically look at the joint positions and angles the rider is achieving on their existing bike as well as their centre of gravity and weight distribution,” says Jesse Jarjour, founder of MyVeloFit. “We then cross-reference that information to their stated riding goals and mobility to make recommendations on what changes (if any) the rider should make to their bike.”
Jarjour says the advantages of a remote AI-enabled bike fit include being able to regularly monitor the fit. “It also means it’s easier if you end up going to an in-person fitter,” Jarjour adds.
Powering training
AI-powered bikes, such as the CAROL and Renpho, use biometric data to personalise our workouts based on power output, goals and cadence.
Renpho, for example, uses AI to generate workout algorithms that personalise the session by automatically adjusting the resistance level. There’s also a power test to measure users’ FTP and identify the right training-intensity zone.
AI is also being used to create virtual routes. For example, Wahoo’s (recently closed) RGT Magic Roads app enabled users to upload their own GPX file from any route in the world and have RGT build an accurate virtual version.
Ruddock says this shared functionality similarities with generative AI, creating novel virtual cycling routes based on an input route or prompt.
“I can see generative AI technologies being more tightly integrated to take this kind of technology a step further, building new and unique virtual worlds that match your training type and intensity based on your current performance and goals,” he says.