The HPC Behind the Product Series
The HPC Behind the Product Series
February 24, 2021
One of my almost-daily routines is riding the Peloton. Besides the smooth ride, engaging classes, and workout options, I appreciate the detailed metrics displayed on the large screen during and after my sessions.
Ever since my first ride, I’ve had a few questions about the engineering behind the Peloton that makes detailed metrics possible:
- How exactly does the Peloton calculate my total output, represented in kilojoules (KJ)?
- What simulations and artificial intelligence were involved in developing the Peloton?
- How does the Peloton know I’m not working as hard, presented to me by way of a lower output metric when I take my hands off the handlebars?
In this blog, we’ll uncover the answers to these questions and with the help of a Multi-Physics and Advanced Simulations Engineer, you’ll learn about the simulations and artificial intelligence that likely went into the development of the Peloton. Although this article was written with the support of a Multi-Physics and Advanced Simulations Engineer, the following HPC examples are assumptions.
The Peloton is one of the first exercise bikes to use machine learning algorithms, creating a superior user experience and providing measurable, meaningful output to riders. Computer modeling and simulation played an essential role in designing and developing the Peloton.
From a hardware perspective, the Peloton bike boasts a welded steel frame that hosts the user interface accessories and the resistance flywheel. A Poly V® power transmission belt drive connects the light-weight pedals to the flywheel.
The frame, flywheel, and pedals work in unison with an array of sensors and intelligent algorithms to provide riders with feedback to maximize their workout results. The design team likely used engineering analysis and verification testing to ensure the moving and stationary parts work well together.
How Does the Peloton Calculate my Total Output?
As I’m riding the Peloton, I see my current output, measured in watts, which tells me how much power I am exerting at that moment. Once I complete my ride, my total output is presented to me in kilojoules (KJ). The harder I work, the higher my KJ, but how exactly does the Peloton calculate my total KJ?
The software tracks my total ride time and uses my average output to convert my effort into calories using the following formula:
For example, over a regular session that lasts 50 minutes (3,000 seconds), a rider delivering an average of 100 watts will output a total of 300 KJ. The average power that a rider delivers depends on the rider’s weight and height, resistance level dialed in from the resistance knob, and ride cadence. A built-in load cell sensor measures the rider’s output which the software converts into calories.
The Simulations Behind the Smooth Ride
Judging from the smooth operation of the Peloton, several computer modeling and simulation methods were employed to optimize the static frame as well as the interaction between the frame and the high-speed moving parts. Implicit finite element methods such as static structural analysis were used to ensure the frame’s robustness, while explicit dynamic methods were utilized to hone-in on the impact of the moving parts on the frame.
Given the size and complexity of the Peloton, high performance computing (HPC) played a crucial role in successfully running those simulations. Complementary to the structural analysis, most of the Peloton’s electronic components, including the touchscreen, have been analyzed using complex electromagnetic/CFD software on the cloud to design the bike’s electronic brain, minimize touchscreen temperature, and avoid throttling of their integrated circuits.
How Does the Peloton Know I’m Not Working as Hard When I Take my Hands off the Handlebars?
It never fails that when I let go of the handlebars, my output metric decreases. I feel like I’m working just as hard as when my hands are on the handlebars because my legs are moving at the same cadence and resistance, but somehow, the Peloton knows that I am not working quite as hard.
The Peloton app uses many artificial intelligence features, including predictive analytics, with its software engine. The general rule is that the rider’s position on the bike dictates the rider’s ability to deliver higher energy output. Sitting on the back of the saddle and holding the handlebars gives the rider the best chance to maximize their output.
Once the rider removes their hands from the bars, the output is most likely decreased due to a change in both body position and how the body is supported by the bike. Machine learning software kernels have built this capability into the Peloton and let the user know that they are not working as hard as they could by displaying a lower KJ output.
Peloton uses a 20-minute average output as the basis to compute and display the functional threshold power (FTP), giving you a personalized understanding of your output numbers. Some other stationary bike brands use handlebar sensors to add additional information to the AI engine, complementing behavioral analysis.
The future is bright for the Peloton and other HPC built home-gym equipment companies. Artificial intelligence combined with customized content can only improve from today and adapt to the ever-evolving customer needs. I can only conclude that we are in the infancy of the tech revolution in the home-gym space. The direction is onwards and upwards from here.
Our blog series, The HPC Behind the Product, highlights the simulations and artificial intelligence involved in the development and design of popular consumer products including the Peloton, YETI tumblers, iRobot vacuum cleaner, the MIRROR, and more.
Disclaimer: We are not affiliated, associated with, or in any way represent Peloton.