Jaden Rand SCC 25-26 Proposal

Problem:

Pacing is integral to performance in a cycling Time Trial, and building a plan of how to pace, or use energy can be difficult, in my experience, especially with limited experience and coaching.

My solution: 

By using simulated neural evolution, I can build a model that takes in a course and generates an optimized pacing plan scaled to a specific rider.

My method:

By building a relatively basic simulation of physics, modelling speed based on rider power output, road gradation, and a few more inputs, as well as a model of rider fatigue, AI can be trained using a model to simulate neural evolution to pace itself on a variety of courses. Results from this training can then be tested via comparison to self-paced efforts and pacing used by professional cyclists.

Possible challenges/generalizations:

Building a model of fatigue may be difficult, as there are limited resources on rider energy reserves and how they are used. Fatigue also varies across different riders, so I may have to generalize or base it off inputted variables such as lactate threshold power or heart rate.