Humon was born after a semester-long market research project at MIT indicated that there was a strong pull from athletes to better understand their bodies so that they could improve their training. Specifically, they asked for a better metric of physical exertion to understand the limits of their body, but without the inconvenience of tests such as lactate threshold or VO2max tests. In this talk, I'll share some of the experiences and findings of the market research, and how the conversations with athletes led to Humon's first product to solve this problem. I'll share information learned along this entrepreneurial journey from market need to product, some of the basic science behind the Humon Hex, describe what muscle oxygenation is and why it is valuable, and some lessons learned along the way. Lastly, I'll discuss some thoughts on the future with Humon.

TBA

TBA

I joined the Stryd team in 2010 at the University of Colorado, starting off by working with all sorts of athletes from college level to olympians. We learned their pain points, and how they are scientifically tested. We also worked with a gait lab at the University of Colorado, capturing motion capture data from various places on the body and seeing if we could do the same thing with our sensor platform as the lab's motion capture system. In the end we found a way to do it accurately, thus bringing the abilities of a motion capture gait lab outdoors into the real world. Gait labs are very useful in diagnosing running deficiencies, but the testing is done on a treadmill for short periods of time. With Stryd we can peer much deeper into the biomechanics of running in real world settings over much longer time periods for each and every step.
 Running turns out to be a very complex sport, and it is no wonder about 70% of runners experience some sort of injury each year. There is a balance of biomechanics and physiology at play where knowledge and feedback on these things amounts to superior training with a proper training load and with a much lower chance of injury. Stryd's unique position is to provide accurate motion capture abilities outside of a gait lab. From this we can quantify subtle changes in running form related to fatigue, performance, and injury. Additionally, by doing high precision motion tracking on the body we can report things that a GPS watch never could. Not only can we peer into each steps motion profile, we can use this to provide superior pace estimation compared to GPS (especially when satellites have a hard time, such as in heavily wooded areas, overcast skies, or cities with large skyscrapers blocking a direct line of sight to the satellite). This holds true in the vertical dimension as well, capturing real time elevation gain and loss, an area that GPS systems struggle with. My talk today will be all about the data we measure, what metrics we report and the physics behind them. I will go over the data we have collected on our athletes for the past 1-2 years, what patterns show up and how we present that data in a meaningful way. I will also talk about some interesting insights we have seen as athletes fatigue as seen from the motion capture data over long runs in the real world.

Industrial requirements on services and products relating to running sports define the limits of the final offering for the customers and the business case. These requirements are important to be considered during the research phase of product development. The industrial requirements comprise technical aspects as well as the handover process between the research phase and the commercialization phase. Example business cases and processes will be shown. Furthermore, the question relating to technology-driven and customer-driven approaches will be discussed as well as the physicist viewpoint of relevant science approaches will be provided.

The foot's stiffness is important for propulsion in walking and running. In this talk, I will discuss past and current understanding of how the uniquely arched structure of the human foot underlies its stiffness. Past work has shown the role of the medial longitudinal arch for stiffness, but has not considered the transverse tarsal arch. Our work shows that the transverse arch is an equally important contributor to foot stiffness. The underlying mechanical principle is demonstrated in stiffening a drooping banknote by slightly curling it in the transverse direction. In concluding my talk, I will briefly discuss other mechanical roles of the foot in running, including mitigating collisions and aiding stability in running.

The athletic performance of a runner depends on many physiological factors such as muscular strength, cardiac efficiency, resistance to acidosis, or the perception and management of pain. Physiology describes the mechanisms involved during physical exertion, including the aerobic and anaerobic mechanisms of energy production that are critical in endurance events, such as running [Billat, 2017]. However, although experimental studies have been able to accurately quantify these mechanisms, the choices that lead a runner to vary his speed are still an active topic of research [Billat, 2000], [Tucker, 2009]. In particular, the "hybrid" character of the runner is well known and has already been considered [Keller, 1974], [Aftalion, 2014] to define a speed profile that optimally solves aerobic and anaerobic mechanisms. Nevertheless, the link with the physiological signals measured in real time, and the individual characteristics of the runners remain difficult to establish. The limitations of the applicability of this work to the practice of sports activity come partly from the fact that the framework used is that of the theory of deterministic optimum control in continuous time. This mechanistic approach, close to physical modeling (Keller) or "classical" physiology [Keener, 2010] is based on the introduction of an evolution equation (ordinary differential equation) of the physiological state, a functional cost of minimize and constraints on controls and state variables. The elicitation necessary to formalize the problem is complex, and incomplete in the best case. Indeed, it has not yet been possible to conclude on the physiological relevance of the framework used, nor on that of the associated hypotheses. A first question that arises is that of the "good" definition of the physiological state (choice of physiological variables to follow: heart rate, oxygen consumption, muscular and biomechanical characteristics, others). In a second step, it is to identify the rule of evolution of this state during a race. Finally, a third question consists in determining the functional minimized by the runner. If the runner tries to minimize his travel time, he can also try to control simultaneously the risk of acidosis, muscle fatigue, ... The terms involved in this functional remain in fact unknown, because the factors involved in the decision variations of speed of a runner are misunderstood, and may be dependent on the physiological characteristics of a person. Finally, feedforward control loops or cognitive aspects are difficult to integrate although they have a significant role in the physiology of exercise [Miall, 1996]. In parallel, the development of many quality sensors for the general public measuring speed, acceleration, heart rate, pace characteristics, ... created a strong expectation of technical and scientific support, from amateurs or professionals. The current existence of substantial databases of physiological signals measured in the field and in real time (during races in particular) makes it possible to consider the construction of approximate models (statistics / machine learning), constructed from data. The challenge is to succeed in building race strategies that improve current race profiles, based on an uncertain knowledge of the rider (ie his physiological state, and his dynamics), and his environment. (the "cost" of a race being unknown and subject of modeling). For this reason, we propose to develop a "Markovian Decision Process" model that better integrates this uncertainty, and to use reinforcement learning techniques to build such strategies, which can be used in real time, and improve with the observation of new races.

The curvilinear relationship between the average speed (v) and the total duration of the race (T) in foot racing (the running curve) is an example of the performance curves describing the maximal effort which can be sustained as a function of T by any animal in any type of activity. A.V. Hill (1886-1977, Physiology or Medicine Nobel Prize, 1922) was probably the first to understand that these curves can be explained on the basis of the mechanisms which supply energy in the muscle, and to recognize that “the records of athletic sports and racing” are the most accurate performance curves available and can help understanding exercise bioenergetics. In 1925 [1], he suggested that the running, cycling, swimming and horse racing curves plotted from world records (WR) can be approximated from the size of anaerobic energy stores (the “O2 debt”) and the maximum output of the aerobic energy source (“maximum O2 intake” i.e., VO2max). These characteristics (noted here A and B) were estimated to be ~15 L of O2 (75 kcal) and 4 L O2/min (20 kcal) for an “average athletic man”. The power output available, expressed in kcal/min is, thus: PT = A/T + B (75/T + 20 e.g., 57.5 and 35 kcal/min or 4025 and 2450 W for T = 2 and 5 min). This equation is the basis for all subsequent hyperbolic models of performance: e.g., Llyod, Wilkie, Ward-Smith: see references in [2], di Prampero [3]). Hill recognized two of the limitations of the model, i.e., that the entire value of A is not available for distances shorter than about ¼ mile; and that PT for distance > 2 miles cannot be fitted to an hyperbola with an horizontal asymptote = B because VO2max cannot be sustained beyond about ~500 sec. The modified hyperbolic model we proposed in 1989 [2] included corrections for these two limitations: the correction suggested by Lloyd for the kinetics of utilization of A at the onset of exercise; and an original correction for the reduction of the value of B beyond 7 min using and index of endurance capability (E). Also included were two corrections for the kinetics of VO2 at the onset of exercise as suggested by Ward-Smith, and for the fact that the entire value of A is not available for T > 7 min. Using this modified hyperbolic model and an equation suggested by di Prampero (see [2]) for estimating PT, we show that the WR (100 m to Marathon) for a given year can be fitted with a very small error, e.g., for the year 1987, 0.67 and 1.08% for men and women with values of A, B (expressed in VO2max), and E (expressed in % VO2max @ Marathon) close to those observed in elite runners: 1691 and 1516 J/kg, 83 and 75 mL O2/kg/min, and 84.1 and 85.7% VO2max in men and women, respectively. The model was used to compute the value of A, VO2max, and E for each year between 1912 and 1987 for men, and 1957 and 1987 for women. In men, the error between estimated and actual WR slowly decreases while both A and VO2max linearly increased with no evidence of levelling off, and E remained stable around ~84 % VO2max @ Marathon. The values of A and VO2max extrapolated for the year 2000 (i.e., 13 years ahead) were 1716 J/kg and 85.2 mL O2/kg/min. Using these values and with E = 84 % VO2max, the WR estimated were very close to the corresponding actual WR in 2000: average error between running time = 0.72%. A similar computation for women also showed a small although slightly larger error: 1.25%. 1. Hill, A.V., The physiological basis of athletic records. The Scientific Monthly, 1925. 21(4): p. 409-428. 2. Peronnet, F. and G. Thibault, Mathematical analysis of running performance and world running records. J Appl Physiol, 1989. 67(1): p. 453-65. 3. Di Prampero, P., et al., Energetics of best performances in middle-distance running. Journal of Applied Physiology, 1993. 74(5): p. 2318-2324.

Our work has focused on developing and elaborating the complexity-loss theory of aging and disease. This brief talk will overview 1) the relevance  of fractals and nonlinear dynamics in medicine; 2) the concept of exercise as an elixir of healthy complexity; and 3) the potential utility of new computational tools in monitoring the “magic bullet” of training and of its off-target effects.

Measurements of physiological variables during exercise and performance evaluations and predictions are important for a fundamental understanding of physiological processes, training and assessment of athletes, and beyond sport in the study of complex physiological response related to aging, muscular structure and cardiovascular health. Models for human running performances of various complexities and underlying principles have been proposed, often employing a combination of data for world record performances and concepts that are not always based on simple principles of human physiology. We present a novel, minimal model for human running performance that follows from a self-consistency relation for the time dependent power output during racing events. The model has a total of four parameters that are not fixed a priori and characterize individual physiological profiles for a runner. The analytic approach presented here is the first to derive the observed logarithmic scaling between world (and other) record running speeds and times from basic principles. Various female and male record performances (world, national) and also personal best performances of individual runners for distances from 800m and to the Marathon are excellently described by our model, with mean absolute errors of (often much) less than 1%. Physiological parameters of our model, as obtained from records and individual runners, are consistent with existing laboratory measurements. The computed maximal power output that can be sustained for a given time describes well existing experimental data for the time to exhaustion dependence of supramaximal oxygen consumption in the anaerobic regime.

The model for running performance presented in the preceeding talk is applied to define and estimate endurance for both the aerobically and anaerobically dominated performances. As an application of the model, we derive personalized training speeds for prescribed duration and intensity. Our findings could be a basis for plethora of further studies including assessment of performance dependence on age, altitude, muscular structure, specialization of athlete, racing strategies, and optimal dosing of recreational exercise.