The somersault is a maneuver in which a performer jumps into the air and rotates through 360 degrees before landing on the ground. The somersault is challenging to execute because the performer has limited control over the body while airborne. Despite conservation of momentum con- straints, limited control of such a maneuver during flight is possible. We experiment with the use of an in flight feedback strategy to control the somer- sault landing of a biped robot. In flight, the robot manipulates its rotation rate by tucking its legs to control the landing attitude. Results from somer- sault experiments in the laboratory provide initial verification of the control strategy.

Control Of A Biped Somersault In 3D

Dynamics and trajectory planning of a planar flipping robot

A type of planar one-legged robot is proposed which, unlike a previous one-legged robot with springy legs, consists of three revolute joints, two links and two feet. In Geng et al. (2001), a cyclic jumping gait has been designed for this robot, which required that angular rate of the knee joint be zero at the instant of takeoff and that the knee joint be locked during flight phases. To eliminate these two requirements that were difficult to implement; a different ballistic flipping gait is designed for the robot in this paper. A complete flipping gait cycle is composed of four phases: two stance phases and two flight phases. During flight phases, torque of the knee joint is zero, the total angular momentum with respect to the center of mass of the system is conserved, rotational motion and translational motion is decoupled from each other. Trajectory planning of the four phases is formulated as a problem of numerical optimization subject to nonlinear constraints such as positive reaction force of ground and finite torque of the joints. Energy injected into the robot in a gait cycle is chosen as a criterion of optimization in the simulation.

Designing ballistic flipping gait for a one-legged robot

Two contributions of the paper are that: (1) a novel manner of one-legged locomotion, ballistic flipping, is proposed; (2) by using the method of a Poincare map, the fixed points corresponding to unstable cyclic flipping gait are found in simulation, then a strategy based on a neural network is presented to stabilize the fixed point. This robot needs no active control during its flight phases. Compared with other one-legged robots that had to regulate their attitudes while in flight, this robot may be energetically economical.

Proposing a novel one-legged robot: control via Poincare map

A novel manner of one-legged bipedal locomotion (i.e., ballistic flip) is proposed in this paper. A flip is composed of three sequential stages: stance phase, flight phase and touchdown. During the stance phase, the two objectives of control, constant locomotive speed and taking off at the end of stance phase, are achieved via a control algorithm based on nonlinear feedback linearization. The flight phase needs no active control. Combining dynamics of the three stages, we construct the Poincare map numerically for a complete flip gait, and then use a gain-tuning strategy to stabilise the fixed point of the Poincare map corresponding to an unstable cyclic flip gait One simulation result is given to verify the effectiveness of this strategy.

http://ieee-ies.org/iecon01/abstracts/IE01A-986521656.pdf

Flip gait synthesis of a biped based on Poincare map

The author studies the dynamics of a planar cat. He then proposes a few strategies that a cat can use to perform a forward or backward somersault. >

Z. Li

Papers

Strategies for biped gymnastics