This paper demonstrates the practical performance of a new theory of balance control that has been shown in simulation to out-perform earlier balance control theories in the sense of allowing the robot to make larger and faster movements while still maintaining its balance. The case studied here is that of a general planar double inverted pendulum, which resembles a legged robot's behaviour when the polygon of support shrinks to a line. The results show the speed and accuracy of the controller, as well as its robustness to external disturbances and slipping during fast movements.

This paper presents the first practical demonstration of a recently developed theory of balance control that aims to achieve high performance in the sense of allowing a robot to make large, fast movements while maintaining its balance on a narrow support. This theory includes a simple method of leaning in anticipation of future motion commands, which is largely responsible for the high performance. The experiments reported here use a robot acting as a reaction wheel pendulum, and they test only the 2-D version of the theory. The results show that the balance controller's performance in practice closely resembles its theoretical performance. This paper also presents a simple yet accurate balance offset observer that measures the difference between true and estimated balanced configurations.

This paper first presents a formula to predict the largest balance disturbance from which a legged robot can recover without taking a step. It then presents an actuator design criterion derived from this formula that maximizes the robot's ability to recover. In this study, it is assumed that the robot is using a single major joint (e.g, a hip joint) to perform its balance recovery movement, and that the actuator consists of an electric motor and reduction gear. It is also assumed that the robot's support polygon is sufficiently small that it can be approximated as a point, and that the balance recovery motion is essentially planar, so that a 2-D analysis remains valid in 3-D. Finally, it is assumed that, for the purpose of studying balance recovery motion, the robot can be approximated by a reaction wheel pendulum. The theory has been tested experimentally on a robot designed to be good at balancing, and was found to agree closely with experimental results.

In this paper we motivate and present our embodied visual programming research plan, aiming at allowing end-users without any technical expertise to define complex robot behaviours. The core idea is to combine visual programming with the sensing and actuation capabilities of robots and with teleoperated demonstrations of what needs to be achieved.