A recent study conducted by University of New Hampshire researchers explored how integrating machine and deep learning techniques can create a standardized system for evaluating rock climbing routes to provide a difficulty grading scale that promotes inclusivity, accuracy and accessibility for all experience levels. The study, published in the journal Frontiers in Sport and Active Living, found that the most successful approach for determining the difficulty of a rock-climbing route used route-centric, natural language processing methods.
As the sport of rock climbing continues to gain popularity and international recognition since making its debut in the 2020 Tokyo Olympics, the demand for a consistent method of determining route difficulty has become increasingly important as there is no official standard. Whereas commercial climbing gyms require objectivity in creating routes that are attainable for climbers of all levels, difficulty scales often rely on personal judgment and remain a subjective process, leading to inconsistencies and bias.
"Rock climbing's popularity as a recreational sport is growing dramatically," says Blaise O'Mara, a graduate student at UNH and lead author of the study. "Reporting the objective grade of a climbing route is critical in the climbing community but the challenge has been in how to set a uniform grade that applies to all skill levels."
Route difficulty relies on multiple factors such as climbing environment, rock hold types and the movements of the climber. The researchers looked at how these factors play a role in the determination of route difficulty. The survey conducted by UNH categorized machine learning techniques into route-centric, climber-centric and path-finding approaches and highlighted the potential use of natural language processing to offer a more objective method for rating route difficulty.
"Through our research, we seek to address how climbing gyms can integrate machine and deep learning systems to streamline route setting and eliminate route difficulty bias," said MD Shaad Mahmud, associate professor of electrical and computer engineering at UNH. "During our study, the route-centric approach focused on analyzing route features such as hold types, movements between holds and sequences, while the climber-centric approach involved using wearable sensors to track metrics like electromyography and acceleration and looked at past climbing performances. The path-finding approach combined qualities from the other approaches. In the end, the most successful of these approaches was the route-centric and path-finding data with natural language processing methods."
The survey found that accuracy and granularity are the key outputs to optimize, and the route-centric method was able to achieve the greatest granularity accuracy of 84.7%. According to UNH researchers, future success in determining rock climbing difficulty in chaotic environments will likely rely on route-centric data extracted with computer vision and then fed through a Natural Language Processing algorithm. Additionally, they expect machine learning and deep learning methods to keep evolving to solve route problems like climbers. With further evolution, these methods may solve the pervading grading bias problem in determining rock climbing route difficulty.
This work was supported by N.H. Agricultural Experiment Station CREATE grant (11HN37).
