← Back to teams list

Team website
Qualification video
Team short paper
Hardware specifications

Software description


Please give a brief summary of your walking algorithm (max. 1000 characters).

We design the foot trajectory of waling robot by using quintic spine. The step frequency can be set manually. Step length is calculated by PID algorithm according to ball distance except for smoothed start and end phases. The maximum step length is restricted to prevent too large step for the robot generated by far ball. Therefore the robot will go to the ball at maximum speed when the ball is far and decelerate when going near the ball. Finally its walk speed decreases to zero when the robot can kick the ball.


Please give a brief summary of your vision algorithm, i.e. how your robots detect balls, field borders, field lines, goalposts and other robots (max. 1000 characters).

We adopt fast multi-target recognition method for humanoid soccer robot. Haar-like features integrated with morphological analysis is employed to construct heat maps, on which the candidate patches supposed containing object in the robot's field of view are roughly selected to form a set of “region of interest (ROI)”. This ROI set is quickly classified to target or non-target regions through a lightweight tiny-dnn convolution neural net-work. Meanwhile, each ROI generated automatically during the robot operation can be directly collected stacked into positive or negative sample sets for the off-line training of the convolution neural network, thus avoiding tremendous labor work and uncertainty deviation of manually tagger of target sample on quantities of pictures. The proposed method is applied to CPU supported SYCU-Legendary humanoid robot, enabling it to recognize football, goalposts and penalty marks within 0.03 seconds.


Please give a brief summary of how your robots localize themselves on the field (max. 1000 characters).

The robot self-positioning model is built by employing particle filter (PF) algorithm. In detail, a set of weighted particles are generated to indicate the probability of the robot being on certain position, expressed by posteriori probability density function. The probability of every weighted particle is calculated iteratively based on the value of the observation model of the robot. After recursive and iterative calculation, the final collection of particles with largest weight represents the most probable position of the robot.


Please give a brief summary of your behavioral architecture and the decision processes of your robots (max. 1000 characters).

In an ordinary mach, our on-site robots share their own localization the ball position to the robot captain who assigns all the robots do what they should. For example, the robot most close to the ball goes to shoot while the one near the goal prepare the next kick and the other robot except the goalkeeper walks to the ball but keep a distance to avoid dismantle with the one near ball. The decision tree is taken for behavioral architecture where the behavior of approaching, goal-Keeper, walking, kicking, etc. can be triggered by corresponding prerequisite.


List your previous participation (including rank) and contribution to the RoboCup community (release of open-source software, datasets, tools etc.)

2018.04(Shaoxing, China): We attended 2018 RoboCup China Open and won the champion by overcoming the vision problems that we never met before. 2018.06(Montreal, Canada): We joined 2018 RoboCup Competition and ranked the sixth place when we deeply realized that we cannot use Rhoban robot as skidedlly as we should so we decided to invite Rhoban team to train us hereafter. 2019.04(Shaoxing, China): We attended 2019 RoboCup China Open where we met ZJUDancer team who is both a strong opponent and a good friend for us. Although we won the final game after intensive fighting, we were convinced by their talent of making the robots better and better during the competition. 2019.07(Sydney, Australia): We took part in 2019 RoboCup Competition and listed the fourth which indicated our improvement compared to the year before. But we were also aware of the prominent progress of our peer teams. We must try our best to promote the robot technology in order to keep up with them for long term.


Please list RoboCup-related papers your team published in 2019.

Wu Fenghua, Li liande, Wang Chengye, Jin Xin, Wang Hao, Yang Zhehai, Yin Jingyao, and Zhang Yuping. Design and Development of Autonomous Soccer Humanoid Robot CU-legendary. Journal of Physics: Conference Series, 2019:1176(3): 266-271. 2018 International Seminar on Science and Engineering Technology, SCSET 2018-Robotics and Artificial Intelli-gence, March 26, 2019. (EI:20191406738791). Fenghua Wu, Tingxue Li, Liande Li, Zhehai Yang, Jingyao Yin, Hao Wang, Xin Jin, Chengye Wang, Yuping Zhang. Robot Subject Establishment through Popularity, Scholar, Research and Employer[J]. Laboratory research and exploration: 2019,38(08):189-196+208. Wu fenghua, Li Liande, Wang Hao, Wang Yueyong, Jiang Jiao, Chen Si. Research on Key Technologies of Humanoid Robot[J]. Artificial Intelligence and Robotics Research, 2017, 06(03): 97-105.