Optimal Pacing of a Cyclist in a Time Trial Based on Individualized Models of Fatigue and Recovery

TL;DR

This study develops individualized models of fatigue and recovery to optimize cyclist pacing in time trials, demonstrating significant time reductions through analytical and numerical methods, including real-time feedback.

Contribution

The paper introduces a novel optimal control framework incorporating personalized fatigue models and mode switching analysis for cyclist pacing optimization.

Findings

24% time reduction in simulations over self-paced performance

Optimal pacing modes are limited to four distinct strategies

Real-time pace feedback reduced trial time by 3%

Abstract

This paper formulates optimal pacing of a cyclist on hilly terrain time-trials as a minimum-time optimal control problem. Maximal power of a cyclist serves as a time-varying constraint and depends on fatigue and recovery which are captured via dynamic models proposed early in the paper. Experimental protocols for identifying the individualized parameters of the proposed fatigue and recovery models are detailed and results for six human subjects are shown. In an analytical treatment via necessary conditions of Pontryagin Minimum Principle, we show that the cyclist's optimal power in a time-trial is limited to only four modes of all-out, coasting, pedaling at a critical power, or constant speed (bang-singular-bang). To determine when to switch between these modes, we resort to numerical solution via dynamic programming. One of the subjects is then simulated on four courses including the 2019 Duathlon National Championship in Greenville, SC. The dynamic programming simulation results show 24% reduction in travel time over experimental results of the self-paced subject who is a competitive amateur cyclist. The paper concludes with description of a pilot lab experiment in which the subject trial time was reduced by 3% when the near-optimal pace was communicated to her in real-time.