Photo credit: www.sciencedaily.com

Advances in Rock Climbing Route Evaluation Through Machine Learning

Researchers from the University of New Hampshire have conducted a comprehensive study examining how machine and deep learning technologies can standardize the grading of rock climbing routes. The findings, which appear in the journal Frontiers in Sport and Active Living, propose a difficulty rating system aimed at enhancing inclusivity, accuracy, and accessibility across various skill levels in the sport.

As rock climbing’s profile rises, particularly following its inclusion in the 2020 Tokyo Olympics, the necessity for a reliable method to assess route difficulty has become more critical. The absence of an official grading standard hampers the creation of routes in commercial climbing gyms, where a consistent approach is essential to accommodate climbers with diverse capabilities. Typically, the subjective nature of current grading systems leads to inconsistencies, making it challenging to provide climbers with a clear expectation of what to anticipate.

Blaise O’Mara, a graduate student at UNH and the primary author of the study, noted, “With the growing popularity of rock climbing as a recreational activity, it’s vital to report objective grades for routes. The hurdle lies in establishing a unified grading system applicable to climbers of all skill levels.”

The evaluation of route difficulty encompasses a variety of elements, including the climbing environment, the types of rock holds, and the physical maneuvers of the climber. The research team aimed to dissect how these variables contribute to overall route assessments. Their survey categorized machine learning methodologies into three strategies: route-centric, climber-centric, and path-finding approaches, emphasizing the effective application of natural language processing to achieve objective evaluations of route difficulty.

MD Shaad Mahmud, an associate professor of electrical and computer engineering at UNH, explained, “Our research addresses how climbing gyms can utilize machine and deep learning systems to refine route setting and reduce bias in difficulty grading. We focused on the route-centric method to analyze various route features such as types of holds, movement transitions, and sequences. The climber-centric method involved the use of wearable technology to collect metrics, including electromyography and acceleration data during climbs. The path-finding approach synthesized elements from both strategies, ultimately leading to the route-centric and path-finding methods with natural language processing as the most effective solution.”

The study revealed that achieving both accuracy and granularity is crucial in this context, with the route-centric method yielding a granularity accuracy of 84.7%. Looking ahead, the researchers predict that advancements in route difficulty assessment, especially in complex environments, will hinge on extracting data through computer vision techniques, subsequently analyzed by natural language processing algorithms. They anticipate that as machine and deep learning technologies develop, they will tackle prevalent challenges in grading biases associated with assessing rock climbing routes.

This pioneering work received support from a CREATE grant through the New Hampshire Agricultural Experiment Station (11HN37).

Source
www.sciencedaily.com