Chenglong Fu

Gait Synthesis and Sensory Control of Stair Climbing for a Humanoid Robot

Stable and robust walking in various environments is one of the most important abilities for a humanoid robot. This paper addresses walking pattern synthesis and sensory feedback control for humanoid stair climbing. The proposed stair-climbing gait is formulated to satisfy the environmental constraint, the kinematic constraint, and the stability constraint; the selection of the gait parameters is formulated as a constrained nonlinear optimization problem. The sensory feedback controller is phase dependent and consists of the torso attitude controller, zero moment point compensator, and impact reducer. The online learning scheme of the proposed feedback controller is based on a policy gradient reinforcement learning method, and the learned controller is robust against external disturbance. The effectiveness of our proposed method was confirmed by walking experiments on a 32-degree-of-freedom humanoid robot.

A simple walking strategy for biped walking based on an intermittent sinusoidal oscillator

This paper presents a control algorithm for biped walking by extension of previous work in the fields of central pattern generator (CPG) and passive walking. The algorithm takes advantage of the passive dynamics of walking, assisting only when necessary with an intermittent sinusoidal oscillator. The parameterized oscillator is used to drive the hip joint; the triggering and ceasing of the oscillator during a walking cycle can be modulated by the sensory feedback. The results from simulation indicate a stable, efficient gait, and robustness against model inaccuracy and environmental variation. We also examine the effects of oscillator parameters and link parameters on the gait, and design a controller to suppress the bifurcation phenomenon based on the error of prior step periods.

A passive walking strategy for a biped robot with a large mass torso by a spring and a damper

This paper presents a control system for a biped robot with a large mass torso by extension of previous work in the field of CPG (Central Pattern Generator). The algorithm takes advantage of the passive dynamics of walking, assisting only when necessary with an intermittent oscillator driving the hip joint. However, a biped robot with a large mass torso cannot walk stably only by the CPG controller and a simple feedback control. We speculate about inherent reasons of falling based on dynamics principles and propose a method to solve this problem. This proposed method is that each leg of the model is connected to a torso by a virtual spring. Furthermore a virtual damper with the variable damper coefficient is attached between a torso and a stance leg. It makes the horizontal moving of the torso easy by shortening the step length and swinging the torso. The results from simulation indicate our controller has a good effect on the biped robot walking with a large mass torso.

Biped blind walking on changing slope with reflex control system

This paper presents a novel reflex control system for passive biped walking on unknown slope varying terrains by extension of previous work in the fields of CPG. The algorithm takes advantage of the passive dynamics of walking, assisting only when necessary with an intermittent oscillator driving the hip joint. We analyze inherent reasons of falling for a biped system based on dynamic principles and assume that human walking relies more on instinct actions within the spinal cord rather than brain. Corresponding falling tendency function is proposed, based on which a reflex controller adjusting output of oscillator is designed. An auxiliary span angle controller is put forward to provide secondary actuations for compensation to improve landing performance. The proposed reflex system requires no prior knowledge of the terrain or tens of hundreds of experiments, which are necessary for machine learning methods. Results of simulations indicate that our reflex controllers are capable of ensuring stable passive blind walking on slope varying terrains without ankle torque.

Proving Asymptotic Stability of Dynamic Walking for a Five-Link Biped Robot with Feet

During the dynamic walking of biped robots, the underactuated rotating DOF emerges between the support foot and the ground. This makes the biped model hybrid and dimension-variant. In this paper, we present the definition of orbit stability for dimension-variant hybrid systems (DVHS). Based on the work of Grizzle et al. (2001), we generalize Poincare theorem to a class of DVHS, and this result is then used to study asymptotically stable dynamic walking for a five-link planar biped robot with flat feet. Time-invariant gait planning and nonlinear control strategy, which is organized around the hybrid zero dynamics of Westervelt et al. (2003), is also introduced to realize dynamic walking with feet. Simulation results indicate that an asymptotically stable limit cycle of dynamic walking is achieved, and the effectiveness of the proposed method is illustrated

Parametric Walking Patterns and Optimum Atlases for Underactuated Biped Robots

This paper addresses parametric walking patterns and optimum design issues for an underactuated biped robot. The walking pattern is curved by impact postures and middle postures. Impact postures are regulated by two parameters and middle postures are selected to adapt the swing foot to negotiate obstacles. To evaluate constraint conditions, stability margins, and walking performances, some indices are defined, and the correlations between these indices and the two parameters are illustrated by the corresponding atlases. The optimum design method which considers multi-criteria is carried out by using these atlases. This method provides not only one optimum result, but also an optimum region which contains all feasible optimum results. Using these atlases presented in this paper one can obtain the optimum result with respect to any object(s)

A walking control strategy combining global sensory reflex and leg synchronization

Biped walking can be regarded as a global limit cycle whose stability is difficult to verify by only local sensory feedback. This paper presents a control strategy combining global sensory reflex and leg synchronization. The inverted pendulum angle is utilized as global motion feedback to ensure global stability, and joint synchronization between legs is designed to stabilize bifurcations. The proposed strategy can achieve a stable gait and stabilize bifurcations. The robustness of this approach was evaluated against external disturbances. Walking experiments of a biped actuated by pneumatic muscles were conducted to confirm the validity of the proposed method. Instead of tracking predetermined trajectories, this method uses sensory reflexes to activate motor neurons and coincides with the biological idea wherein inessential degrees-of-freedom are barely controlled rather than strictly controlled.

Bipedal walking with toe-off, heel-strike and compliance with external disturbances

Both disturbance rejection and human-like motions, like toe-off and heel-strike, are important for a biped robot to enhance its performance. However, the required motions for them may influence each other, which is why few studies consider them simultaneously. This paper presents a method to realize stable walking with toe-off and heel-strike even when the robot experiences disturbances. We propose a walking controller which can constrain the desired ground reaction force. On the one hand, the controller can adjust the torso acceleration to make the supporting leg compliant with the external disturbances on the torso. On the other hand, it can rotate the supporting foot by adjusting the ZMP to an appropriate location. The linear inverted pendulum is utilized to generate the CoM trajectory and the foot placement. Meanwhile, its ZMP can be predefined as required so that the toe-off and heel-strike can be achieved by the controller. The effectiveness of the proposed method is demonstrated by simulations.

The key parameter selection in design of an active electrical transfemoral prosthesis

This paper presents an electrical powered transfemoral prosthesis. The knee joint is actuated by a motor-driven ball screw assembly and the ankle joint is motioned by a series elastic actuator. An undirectional spring is attached paralleled with the SEA to store energy during dorsiflexion and bias load of ankle torque. To mimic biomechanical gait data and make the prosthesis compact and anthropomorphic, some key parameters are calculated to achieve optimal solution. These parameters include the length of crank rocker of knee and ankle joint, the stiffness of series and parallel spring and the dimension of the series leaf spring. The analysis procedure is demonstrated in this paper.

Planning and Control for THBIP-I Humanoid Robot

This paper presents the gait planning method and control strategy for THBIP-I (Tsinghua biped humanoid robot) to realize stable walking. This robot includes a head, a trunk, two arms, two legs, and two feet, totally 32 DOF (degree of freedom). This project is involved for five years and aimed to realize stable walking in various environments for a self-contained humanoid robot. Firstly, the design issues including mechanical structure and control architecture of the robot are prescribed. Secondly, gait generation method based on optimization of the main support leg is presented, and control strategy composed of local joint controller and sensory feedback controller is also illustrated. The walking experiments show that the robot has the abilities of stable walking on the ground, turning in any direction, and climbing up/down stairs