US 12,461,545 B2
Intelligent learning and adjustment system for tennis training robot
Li Wang, Jiangsu (CN); and Yong Zhang, Jiangsu (CN)
Assigned to POTENT SPORTS & TECH CO., LTD., Jiangsu (CN)
Filed by POTENT SPORTS & TECH CO., LTD, Jiangsu (CN)
Filed on Apr. 10, 2024, as Appl. No. 18/632,111.
Claims priority of application No. 202410073488.4 (CN), filed on Jan. 18, 2024.
Prior Publication US 2025/0238037 A1, Jul. 24, 2025
Int. Cl. G06V 10/82 (2022.01); G05D 1/656 (2024.01); G06V 20/56 (2022.01); G06V 20/64 (2022.01); G06V 40/20 (2022.01); G05D 101/15 (2024.01); G05D 101/20 (2024.01); G05D 105/60 (2024.01); G05D 111/10 (2024.01)
CPC G05D 1/656 (2024.01) [G06V 10/82 (2022.01); G06V 20/56 (2022.01); G06V 20/64 (2022.01); G06V 40/23 (2022.01); G05D 2101/15 (2024.01); G05D 2101/20 (2024.01); G05D 2105/60 (2024.01); G05D 2111/10 (2024.01)] 5 Claims
OG exemplary drawing
 
1. An intelligent learning and adjustment system for a tennis training robot, comprising an image recognition system, an algorithm model, a back-end processing platform, and an optimization model;
the image recognition system is configured to capture data of tennis balls, such as three-dimensional coordinates, speeds, and speed directions of a plurality of points, and the algorithm model is configured to calculate trajectories of incoming balls according to the captured data to predict placements of the incoming balls;
the algorithm model has a plurality of algorithm layers, and performs predication and classification according to features of tennis image data inputted by the image recognition system;
error loss values between prediction data and true data are calculated by a calculation model according to data of the placements of the incoming balls and athlete levels obtained through transformation operation of data at an algorithm layer, on-site subsequent data received by the back-end processing platform from a main controller, and true values of data from a mobile phone APP, and the back-end processing platform feeds back the error loss values to the optimization model for data optimization; and
the optimization model expands and improves a database according to accumulative data of ball hitting of athletes and feedback on whether predictions are accurate, and provides model training and evaluation functions, such that the tennis training robot is capable of independently evaluating the performance of the athletes on tasks, and the algorithm is optimized based on the extracted evaluation features.