| Summary: | 碩士 === 國立臺灣大學 === 機械工程學研究所 === 105 === This research is dedicated to improve performance of self-made multi-axis force/torque sensor embedded on hip actuator of hexapod robot and apply it to control dynamic gaits of robot. By utilizing finite element analysis to help us understanding mechanism under force/torque sensor stress/strain behavior, new force/torque sensor design produces larger strain signal under same load than last version. Strain gauges temperature drift problem is solved and measured signal is larger by changing quarter bridge to half bridge design. A novel motor housing method is applied to stop direct connection between self-made sensor and motor, making sure that electrical voltage potential of motor would not influence the measuring signal of half bridge. It also decrease possibility of reny screws’ fracture. Overall, these improvements reduce the calibration error of force/torque sensor from 10N maximum to 2N. Hence, measurement values of self-made force/torque sensor can be used as feedback signal to control dynamic gaits of robot.
In simulation, HybridR-SLIP model is proposed, which use PR-SLIP model as foundation, adding ground reaction force feedback to achieve force control. Hip trajectory and ground reaction force profile in R-SLIP model are used as position control target and force control target. After calculating torque requirements in each controller, hybrid controller sums up position and force control results to control hip DoF of the model. Given a deviation in initial touchdown condition, force and position hybrid control making model converging to passive dynamics of target fixed points faster is verified in simulation.
Design and manufacturing Hexapod FROHex, which is more rigid than TWIX, reduces the influence of strain signal of force/torque sensor when deformation of robot is occurred. Results of fixed points in R-SLIP model are used to induce passive dynamics of FROHex. It exists about 100N difference between FROHex and R-SLIP model in high touchdown speed angle fixed point targets. Through force and position hybrid control, measuring force profiles are closer to those of R-SLIP model. Because of smaller difference between dynamics of robot and real passive dynamics of circular legs, electrical consumption are reduced under hybrid control strategy.
In simulation, HybridR-SLIP shows greater converge ability in transient state. In experiments, force feedback control strategy induced passive dynamics in force profile. Force and position hybrid control is verified to improve dynamics of robot in both way.
|
|---|
