There exists a class of two-legged machines for which walking is a natural dynamic mode. Once started on a shallow slope, a machine of this class will settle into a steady gait quite comparable to ...

/pdf/passive-dynamic-walking-49x6lqysvt.pdf

Passive dynamic walking

Virtual model control is a motion control framework that uses virtual components to create virtual forces generated when the virtual components interact with a robot system. An algorithm derived based on the virtual model control framework is applied to a physical planar bipedal robot. It uses a simple set of virtual components that allows the robot to walk successfully over level terrain. This paper also describes how the algorithm can be augmented for rough terrain walking based on geometric consideration. The resulting algorithm is very simple and does not require the biped to have an extensive sensory system. The robot does not know the slope gradients and transition locations in advance. The ground is detected using foot contact switches. Using the algorithm, we have successfully compelled a simulated seven-link planar biped to walk blindly up and down slopes and over rolling terrain.

/pdf/virtual-model-control-an-intuitive-approach-for-bipedal-6nf8n3vn8a.pdf

Virtual Model Control: An Intuitive Approach for Bipedal Locomotion:

To reduce the complex walking dynamics of a biped, a particular class of trajectories of an ideal biped model where the center of gravity of the body moves horizontally and the horizontal motion of the center of gravity can be expressed by a simple linear differential equation is introduced. The authors coin the phrase 'potential energy conserving orbit' to describe this class of trajectories. Based on these properties, control laws were formulated for walk initiation, walk continuation, and walk termination. The walking motion is controlled by support leg exchange. Robust realization of the walking control is also considered. An experimental walking machine was designed as a nearly ideal biped model. To make the legs lighter, four DC motors were mounted in the body, and the legs are parallel link structures. The results of the experiment describe five steps of dynamic walking including walk initiation. >

Dynamic walking control of a biped robot along a potential energy conserving orbit

The authors aimed to realize smooth 3D biped walking in a robot through control based on information obtained from various sensors. They employed a method to control walking by dividing it into motions in the sagittal plane and in the lateral plane. They treated motion in the lateral plane as a regulator problem with two equilibrium states. They also used relatively low gain feedback coefficients obtained from the optimal regulator theory. For motion in the sagittal plane, they put the body speed close to the smooth speed function given in advance by controlling the ankle torque. The effectiveness of the proposed control method was examined by computer simulation and proved by experiments with out BLR-G2 walking robot. The BLR-G2 is equipped with foot pressure and ankle torque sensors to provide information about the condition of contact with the floor. The sole and ankle driving actuators undergo force/torque feedback control based on the sensor information. These contributed toward realizing smooth walking with the sole firmly gripping the floor.

Sensor-based control of a nine-link biped

We describe an evolutionary approach to the control problem of bipedal walking. Using a full rigid-body simulation of a biped, it was possible to evolve recurrent neural networks that controlled stable straight-line walking on a planar surface. No proprioceptive information was necessary in order to achieve this task. Furthermore, simple sensory input to locate a sound source was integrated to achieve directional walking. To our knowledge, this is the first work that demonstrates the application of evolutionary optimization to 3D physically simulated biped locomotion.

/pdf/evolution-of-central-pattern-generators-for-bipedal-walking-gy1rkv2pol.pdf

Evolution of central pattern generators for bipedal walking in a real-time physics environment

The purpose of this paper is to propose a new controlling method of the biped locomotion composed of seven links and to show simulated and experimental data of the locomotive motion of the biped called CW-1 built in our laboratory. The control method depends mainly on the linear optimal state feedback regulator, which stabilizes the stance supported by both legs as a commanded attitude. As a result the biped successfully walks any pre-determined steps such that one step takes only one second.

Toru Yamaguchi

Papers

Realization of a high speed biped using modern control theory