Ground reaction force (GRF) plays an integral role in legged robots to control interaction with the ground. However, most techniques in whole-body controller for quadrupedal robots do not explicitly take into account actual torque or force in their control loops and instead use feed-forward force to generate joint torque at every time step. In this paper, we present a closed-loop whole-body controller using the actual joint torque feedback, which regulates angular momentum of the center of mass (CoM) for quadrupedal locomotion. Using the torque measured from each torque sensor and the torque by solving the inverse dynamics, we can compute the external joint torque induced by the contact with the ground. To fully use the computed joint torque, we discuss a feasible approach and whole-body control criterion for quadrupedal robots that have constrained support polygons because of their point-feet and certain gaits using two or less legs in contact. Based on the approach, we generate a centroidal moment pivot trajectory considering the leg dynamics, linear translation, and angular rotation of the CoM, which can stabilize the robot‘s balance by using the actual angular momentum rate change transformed from the measured joint torque. In addition, a push recovery strategy based on capture point dynamics derived from linear momentum and a foothold generation method are integrated into the controller. The proposed controller is tested on a quadrupedal robot, called AiDIN-VI, that has a torque sensor at each joint. The proposed whole-body controller enables the robot to demonstrate several gait types such as trotting, pacing, jumping, and walking on various environments, and locomotive abilities under external pushes are verified.

地面反作用力 (GRF) 在腿式机器人中起着不可或缺的作用，以控制与地面的交互。然而，四足机器人全身控制器中的大多数技术并没有明确考虑其控制回路中的实际扭矩或力，而是在每个时间步使用前馈力来产生关节扭矩。在本文中，我们提出了一种使用实际关节扭矩反馈的闭环全身控制器，该控制器调节四足运动的质心 (CoM) 角动量。使用从每个扭矩传感器测量的扭矩和通过求解逆动力学的扭矩，我们可以计算由与地面接触引起的外部关节扭矩。为了充分利用计算出的关节扭矩，我们讨论了四足机器人的可行方法和全身控制标准，因为它们的点脚和某些步态使用两条或更少的腿接触，因此具有约束支撑多边形。基于该方法，我们生成了一个考虑 CoM 的腿部动力学、线性平移和角旋转的质心力矩枢轴轨迹，它可以通过使用从测量的关节扭矩转换而来的实际角动量率变化来稳定机器人的平衡。此外，基于从线性动量导出的捕获点动力学和立足点生成方法的推动恢复策略被集成到控制器中。建议的控制器在名为 AiDIN-VI 的四足机器人上进行测试，该机器人在每个关节处都有一个扭矩传感器。