Conceptual Design andImplementation of ArmWrestling Robot

Inthispaper, we develop a novelrobotic arm wrestling systemintegrated withmechanical arm,elbow/wrist force sensors, servomotor,encoder, 3-DMEMS accelerometer, andUSBcamera. Thearmwrestling robot(AWR)isintended to playarm wrestling gamewithrealhuman on a tablefor entertainment. Thedesigning scenario oftheprototype model's hardware isperformed. Elbow/wrist force sensors, asacrucial device intheforcesensing system, aredescribed indetails. Software isdeveloped fordevice drivenandinterface. The surface electromyographic (EMG)signals fromtheupperlimb aresampled whenarealplayer competes withtheforce testing system. Byusing themethodofwavelet packet transformation (WPT),thehigh-frequency noises canbeeliminated effectively and thecharacteristics ofEMG signals canbe extracted. Artificial neural networkisadopted toestimate theelbowjoint torque. Theeffectiveness ofthehumanoid algorithm using torque control estimated viaWRT andneural network isconfirmed by experiments. IndexTerms-AWR;EMG;WPT;Neural networks