Fig 5 - uploaded by Eiichi Yoshida
Content may be subject to copyright.
Source publication
![Fig. 1. Definition of variables [7]](profile/Eiichi-Yoshida-2/publication/268804753/figure/fig1/AS:578702585352192@1514984856165/Definition-of-variables-7_Q320.jpg)
This paper describes a generalized approach for compensating just the required yaw moment of a humanoid robot about the Zero Moment Point (ZMP) while performing an arbitrary motion, in order to prevent unwanted / unexpected yaw rotations. This is done by modifying the motion of any set of joints with low priority tasks that can be arbitrarily selec...
Contexts in source publication
Context 1
... proposed high priority tasks for this robot is described as follows: starting from the position (−1, 1) m the robot has to approach a ball positioned at (0, 0) m by following the footsteps specified in Fig. 5, which also includes the kicking phase. The desired trajectory for the ZMP is also shown in Fig. 5, as well as the one for the projection of the waist position that stabilizes the robot. Both, the kicking motion and the stabilization process, are calculated as explained in [9]. Also, as a part of the high priority tasks it was decided ...
Context 2
... proposed high priority tasks for this robot is described as follows: starting from the position (−1, 1) m the robot has to approach a ball positioned at (0, 0) m by following the footsteps specified in Fig. 5, which also includes the kicking phase. The desired trajectory for the ZMP is also shown in Fig. 5, as well as the one for the projection of the waist position that stabilizes the robot. Both, the kicking motion and the stabilization process, are calculated as explained in [9]. Also, as a part of the high priority tasks it was decided for the robot to use the yaw-axis joint of its neck to look towards to the position of the ...
Context 3
... yaw moment measured by the force/moment sensors during the simulation (and filtered because of the noise), when no compensation is performed and when it is, are shown in Fig. 15. For this case a coefficient of static friction µ s = 0.4 was considered to prevent the robot from falling in any case and get a clear comparison. This time τ th p,z = 2 Nm, a threshold that is almost maintained as desired, except for some ...
Citations
... Additionally, another main contribution of this thesis was the generation of a whole-body motion capable of preventing undesired rotations between the feet and the ground. These rotations deviate the robot from its desired orientation, leading to inaccuracies for the planned pedipulation motion [39] [40]. ...
... The final version of the algorithm including the drift compensation is depicted in Figure 5.5. This one was implemented as an online closed-loop compensator that takes as inputs the readings of the force / moment sensors placed at the feet of the humanoid robot, calculates the induced yaw moment and generates the appropriate joint motion compensation [40]. Case study: free kick in soccer Let us, as a way to exemplify the developed theory, introduce one case study to be analyzed in detail: the free kick in soccer. ...
This thesis addresses the problem of impulsive manipulation of arbitrary rigid objects by using the lower extremities of a humanoid robot, action that is termed as impulsive pedipulation. The objective of the impulsive pedipulation approach explained in this thesis consists on driving an object to a specified 3D position in space, along with some desirable motion characteristics. This shall be done by using any link belonging to the lower extremities of the humanoid robot in order to exert a proper impact on the object. This impact must be capable of transmitting the required momentum to the object for it to move freely in space and reach the desired goal position. By considering to use the lower links as an alternative end effector and being able to succeed in this
kind of tasks, the ``manipulation'' capabilities of the robot would be extended without the consideration of additional hardware.
However, this is not a straightforward problem. First of all, the mechanics of the impact are complex, even if we assume that the objects are rigid. The inertial parameters of the bodies (or systems of bodies) that enter into contact, the topological structure of their surface, their rugosity and the loss of energy during the impact are all entangled during the impact process, for which finding its inverse problem is required. Moreover, exerting an impulse by using the lower extremities of the robot requires to take into account the complex dynamics of the humanoid robot which, besides performing the pedipulation task, has to keep stability, given the presence of under-actuation and unilateral constraints between its feet and the floor.
This thesis solves this problem by first finding a solution for the inverse motion problem; that is, finding the initial conditions for the motion required to achieve the goal. Then, by proposing some conditions to perform the pedipulation such that the problem get simplified, the inverse problem of the impact is solved, resulting in a required approaching velocity of impact that the robot has to fulfill by means of an appropriate trajectory of one of its lower extremities. Having done this, the high redundancy of the humanoid robot is used to compensate the dynamical effects of this required motion, in order to prevent not only its fall, but also any undesired slippage. Slippages cause rotation about the vertical axis that deviate the robot from its assumed orientation. This compensation is done by using a novel method proposed in this thesis, which uses a synergetic motion of any set of selected ``free'' joints.
The whole pedipulation algorithm is then validated by means of simulations and experiments on a real humanoid robot. This is done by choosing a simple pedipulation task easy to describe that includes the necessary complexities to show the applicability of the algorithm: the free kick in soccer, which requires to drive a ball to a desired 3D goal position.
This simple task was, however, not easy to simulate at first. The simulation platform was not able to properly simulate the dynamics of a spherical object because it was by default treated as a polygonal mesh, resulting in an unnatural bouncing and rolling behavior. It was then required to improve the collision detection system of the dynamics simulation engine, by using an algorithm also described in this thesis.
