Actual Zero Moment Point Research Articles

A Humanoid Robot Gait Planning and Its Stability Validation

Gait planning based on linear inverted pendulum (LIPM) on structured road surface can be quickly generated because of the simple model and definite physical meaning. However, over-simplifi- cation of the model and discontents of zero velocity and acceleration boundary conditions when robot starts and stops walking lead to obvious difference between the model and the real robot. In this paper, parameterized gait is planned and trajectories’ smoothness of each joint angle and centroid are ensured using the 3-D LIPM theory. Static walking method is used to satisfy zero velocity and acceleration boundary conditions. Besides, a multi-link model is built to validate the stability. Simulation experiments show that: despite of some deviation from the theoretical solution, the actual zero-moment point (ZMP) is still within the support polygon, and the robot walks steadily. In consequence, the rationality and validity of model simplification of LIPM is demonstrated.

Balancing Control of AIT Leg Exoskeleton Using ZMP based FLC

This paper is focused on the use of Zero Moment Point (ZMP) concept for balancing control of the Asian Institute of Technology Leg EXoskeleton-I (ALEX-I). ALEX-I has been developed to assist patients who suffer from paraplegia or immobility due to the loss of power on lower limbs. The balanced posture set-points (joint trajectories) under ZMP criterion are generated offline. The ZMP based set points are provided as the desired postures to ALEX-I. Fuzzy Logic Controller (FLC) determines the modified set points based on postures balancing sensed by loadcells. Ground Contact Point (GCP) is used to find the “ZMP-like in real time”. GCP data is obtained by placing 4 loadcells forming a force plate on each foot of ALEX-I. This GCP data is then compared with the reference ZMP. Uncertainties of the model parameters, backlash, and joint tolerance are considered as disturbance. The differences of ZMP and GCP on x-z plane are used as the inputs to the FLC. The 4 outputs from FLC are the compensated angles of left and right ankles joints in roll and pitch axes that make the actual ZMP locate in the convex hull of the supporting area.

Online Walking Control System for Humanoids with Short Cycle Pattern Generation

The present paper presents an online walking control system that frequently generates and updates dynamically stable motion patterns with a cycle time of 20 ms. We show that frequently updating the motion pattern contributes to maintaining long-term balance while performing online walking control. In addition, the system enables a robot to respond quickly to changes in the commanded walking direction. Using preview control theory, we generate dynamically stable walking patterns. We propose a method to adjust the future desired zero moment point (ZMP) by modifying the foot landing position in order to maintain the dynamic balance of the generated motion pattern. This technique can be used to filter input commands that would result in sudden changes to the foot landing position, which would result in dynamic instability. The method is also used to compensate for errors between the actual and desired ZMP due to disturbances encountered while walking. We also present an extension of the short cycle pattern generation method that can accommodate external forces measured online. Experimental results for activities such as pushing a table are demonstrated on the full-size humanoid HRP-2 to evaluate the performance of the proposed walking control system.

Recurrent neural network approaches for biped walking robot based on zero-moment point criterion

The main objective of this paper is to use a recurrent neural network (RNN) to determine the trunk motion for a biped-walking machine, based on the zero-moment point (ZMP) criterion. ZMP criterion can be used to plan a stable gait for a biped-walking machine that has a trunk (inverted pendulum). So, a RNN is trained to determine a compensative trunk motion that makes the actual ZMP get closer to the planned ZMP. In this context, an identification scheme is presented to obtain the vector of parameters of the RNN. A first order standard back-propagation with momentum (BPM) is used to adjust free parameters for the network. Artificial neural network brings up important features for function approximation. This was the main reason to use an RNN to determine the trunk motion. The proposed scheme is simulated on a 10-degree-of-freedom biped robot. The results confirm the convergence of the proposed scheme, proving this is a new way to solve this classical problem in the biped-walking machine area.