本論文主要是完成一個人形機器人自動學習投籃力道的籃球訓練平台。本論文依據FIRA HuroCup競賽中籃球項目為主要需求，競賽場地的架構為一個長120公分、寬120公分的範圍內，有一個籃球的半場環境、一個高30公分的紅色籃框、一顆用來替代籃球的乒乓球以及一台視覺全自主並投乒乓球的人形機器人。本論文首先架設一個120cm(長)*120cm(寬)*160cm(高)的平台，並在平台的上方及側面各裝置一個攝影機。機器人所投出的球是透過兩個攝影機以曲線擬合的方式來定位出球在三維空間中的位置。機器人內架設一個投球距離與力道的一輸入一輸出模糊控制器，以決定機器人的投球力道。為了找出適合的模糊控制器參數，本論文以DNA演算法來做系統的參數最佳化。DNA演算法包含了複製、交配、突變、酶與病毒等幾項過程。在引進最佳化演算法後機器人可以由不斷投球的過程中自動找出適合的模糊控制器。最後，本論文透過實驗的數據模擬以及實際的機器人投球訓練來證明此系統的有效性。 A basketball training platform for the shooting ability of small-sized humanoid robots is presented in the thesis. The basketball scene is based on FIRA HuroCup competition. A vision-based autonomous robot needs to throw a ball into a 30cm high basket. A table tennis ball is substituted for the basketball. In order to setup the platform, A 120cm x 120cm x 160cm platform is built up by aluminum. Two cameras are mounted on the top and side of the platform respectively. These two cameras locate the ball position based on curve fitting. A one input and one output Fuzzy system for determining the strength of the throwing is presented in this thesis. In order to find out the best parameter sets of the proposed Fuzzy system, DNA algorithm is applied in this thesis. DNA is an optimization algorithm including, reproduce, crossover, mutation and enzyme and virus. The proposed system is able to monitor the throwing performance when the small-sized humanoid robot is throwing the ball. According to the real experiment data, the robot is able to adjust the parameter sets of the proposed Fuzzy system. The simulation results and experimental data show the efficiency of the proposed method.
