Bounding Gait Simulation

We have made a simple simulation of our quadruped in the open source robotics simulator V-REP. This simulation includes realistic leg, body, and motor masses based on the Lego prototype that we made. The springs on the legs were simulated as a prismatic joints with position control enabled to regulate the force according to a proportional parameter. In this simulation a simple bounding gait, or a gait where pairs of legs in the front and hind move together, was selected. In this gait, the hip actuators were made to follow a sinusoidal oscillation and a phase shift was applied at the hind legs. The amplitude, frequency, and phase difference between the front and hind legs was made to be adjustable through a slider control.  The goal of this simple simulation was to verify that turning could be attained by decreasing the amplitude of leg motion on one side of the robot and increasing it on the other. Previous work by Iida et. al, that this project is based on, did not consider turning behavior. Work done with the Scout II robot indicated that turning with a bounding gait could be attained by increasing and decreasing torque applied to legs on opposite sides of the body. However, Scout II used different legs than the current design and Scout II used torque control instead of position control, as is the case when linkages are used. Thus verification was needed to determine if turning was possible. This turned out to work, however, at extreme amplitudes the robot became unstable and flipped over.

Trot Gait Simulation

We also tried simulating the robot with a trotting gait or a gait where legs diagonal to each other move in unison. Trot gaits tend to be faster and more efficient than bounding gaits, but they have  not been attempted with our leg configuration and turning approach. In the simulation this was done by putting the front-right and hind left legs in phase with each other and having the remaining legs move with a phase difference with respect to the former.. The trot gait was found to produce stable walking behaviour, but when a turn was attempted, the robot fell over. So we will use a bounding gait for our final robot.

Quadrupedal Klann Linkage Design

We considered using a Klann linkage combined with a non-linear spring instead of sinousoidally driven links to walk, because a klann linkage generates a nice half-ellipsoidal path.  Recent work with quadruped robots has shown that by having the legs of a quadruped robot follow a half-ellipsoid path and adjusting the 'stiffness' of the legs through this path to stiffen as they touch the ground highly stable walking can be obtained. We simulated a quadruped with klann linkages and demonstrated stable(albeit bumpy) walking even without nonlinear springs. However, when we attempted to turn, the robot fell over, so we decided not to use it.