July 15, 2013
Updated energy models
With the release of dacadoo 2.5 looming right behind the corner, our Head of Digital Marketing, Yago Veith, sat together with André Naef who is Chief Technical Officer at dacadoo to talk about changes in the energy models that are part of the upcoming release.
Yago Veith: Thanks for taking the time for this interview. First off, for readers not familiar with energy models, what are these models about?
André Naef: The energy models in dacadoo are about estimating the calories burned during a workout. When one tracks a workout with the dacadoo tracker, it is the energy models that calculate the kcals shown on the workout screen. To do their job, these models consider various inputs from the workout.
Yago: What inputs in particular?
André: Heart rate and weight would be common inputs. Beyond that, the inputs depend on the workout activity. Among others, dacadoo considers duration, distance, slope, altitude and velocity in the context of various activities.
Yago: If heart rate is universal, why not use heart rate exclusively?
André: No all of our users can or want to wear a heart rate monitor for each of their workouts. For this reason, it is important that we can estimate the calories burned even if heart rate is absent. Also, it is important to understand that while heart rate is a very valuable input and a reliable model can be built on top of heart rate, this is still only one model. There is no such thing as a “gold standard” model. Having more than one model and being able to fuse them generally increases the accuracy of the energy estimation.
Yago: So this is the reason why non heart rate models are important. What is new with regard to those models in release 2.5?
André: For cyclists and mountain bikers, we have extended the energy models to include kinetic energy. This is of course a term from physics that refers to the energy stored in the rider and the bike while they are in motion. It takes quite a bit of energy to accelerate a bike and its rider to a particular speed. That energy is typically lost when the bike is slowed down by breaking. And the energy must be spent again to speed up the bike anew. So, this is why cyclists do not really like to stop. Now, for extended cycling on the country side without slowing down, kinetic energy does not play a big role. However, if commuting in the city with frequent stops at traffic lights, kinetic energy can be a considerable energy sink. For such workouts, both in cycling and mountain biking, dacadoo 2.5 will increase the accuracy of the estimation.
Yago: Anything else for cycling and mountain biking?
André: Yes. When cycling or mountain biking at high speeds, the wind resistance quickly becomes the dominant factor. We have updated our models to make more flexible and more realistic assumptions about the wind resistance experienced by the rider.
Yago: Let’s go to other fitness activities.
André: dacadoo 2.5 includes updates for running, walking and mountain hiking. For running, our model has been updated to more accurately estimate energy spent when running on steep uphill and downhill slopes, and users should generally see a somewhat higher energy estimation when doing runs with that profile. More significant changes have been applied to the walking model. Our updated walking model considers the velocity of walking in addition to the slope and distance. Users taking brisk walks should see considerably higher energy estimation.
Yago: Why is that?
André: You could say that walking at a higher speed is less efficient. One gets to the goal faster, at the price of spending more energy.
Yago: What is the story with mountain hiking?
André: Our mountain hiking model now considers the distance, slope and velocity, similar to the walking model. This should provide for significantly more accurate energy estimation.
Yago: These are interesting updates for popular fitness activities. Anything else for release 2.5?
André: Finally, we have also updated the swimming models. The updated models consider the velocity of swimming in addition to the duration. The velocity is typically derived by users entering both the duration and the distance they have swum with a particular stroke.
Yago: Why does velocity matter?
André: This is due to the drag caused by the water that increases with velocity. You need to spend considerably more energy to swim at a higher speed.
Yago: So, these changes improve the accuracy of the energy estimation for swimmers?
André: Generally, yes. But in all fairness, it has to be noted that skill plays a hugely important role when estimating calories burned with swimming. If one compares a skilled swimmer with an unskilled swimmer, the energy spent can differ by factor of 2 to 3. This is huge.
Yago: What does dacadoo do about this?
André: Our models currently assume that a swimmer is reasonably skilled. Or to put this another way, chances are that someone who is just beginning with swimming spends considerably more energy than what dacadoo estimates.
Yago: Room for improvement in a future dacadoo release?
André: Definitely. While the 2.5 energy models are a good improvement, working on these models is an ongoing process.
Yago: What models would you like to work on particularly?
André: Alpine skiing, nordic skiing and inline skating are high up on our list. And of course swimming may be revisited in the future to factor in the skill of each swimmer.
Yago: We are looking forward to that progress. Thanks for the conversation, André.