Akademik Veri Yönetim Sistemi
10th MathSport International Conference, Budapest, Macaristan, 26 Haziran 2023, ss.70-75
Bouldering is a well-balanced sport between physical, technical and mental skill requirements as it tests decision making and problem-solving abilities of athletes alongside their physical fitness. In bouldering competitions, each climbing route is a novel decision problem that an athlete needs to solve in a limited amount of time. The athlete needs to plan the sequence of movements for a route they see for the first time before starting to climb. This study presents a goal-based climbing agent that learns from the previous solutions it observes to plan the sequence of actions for novel bouldering problems it encounters. The agent learns the cost of their movements from a dataset of the difficulty estimations and solutions provided by expert climbers and uses this information to solve the novel climbing problems it encounters. It also uses expert knowledge to constrain the learning and state space.