A motion planner of emergency stop must make an operating humanoid robot to a stationary state from the emergency signal. It plays an important role in prevention of falling over because humanoid robots fall over easily. Immediately after the emergency signal, the emergency stop motion must be generated in real-time. We model a humanoid robot as a simple dynamic system consists of the ZMP (Zero-Moment Point) and the COG (Center of Gravity) as its states. The emergency stop motion is generated by a state feedback. We determine optimal feedback gains in terms of the initial conditions and the pole assignment. The proposed method realizes a reliable emergency stop with low computational cost. Furthermore, it can easily predict the possibility of the successful emergency stop at any time. The validity of the proposed method is confirmed by some experiments using humanoid robot HRP-2.
We investigated and identified the conditions necessary for asymptotically stable gait generation in compass-like biped robots from the mechanical energy balance point of view. The equilibrium point at impact in a dynamic gait is uniquely determined by two constraint conditions: keeping the restored mechanical energy constant and settling the relative hip-joint angle to the desired value just before impact. The generated bipedal gait with these two constraints then becomes asymptotically stable around the equilibrium point, as shown by a simple recurrence formula of the kinetic energy just before impact. We verified this stable gait generation method using numerical simulation of virtual passive dynamic walking. The result were compared with those for a rimless wheel and an inherent stability principle was derived. Furthermore, we derived a robust control law using a reference mechanical energy trajectory and demonstrated its effectiveness numerically. Finally, we discuss sufficient conditions for the gait stability in more general cases using an index function.
In this paper, we propose new asymmetric robotic catapults based on the closed elastica. The conventional robotic catapults based on the closed elastica the authors developed are the robotic elements for generating impulsive motions by utilizing the snap-through buckling. In a typical closed elastic catapult, the two ends of an elastic strip are fixed to a passive rotational joint and an active rotational joint, respectively. Here, we found that we could obtain higher impulsive motions by adding the range limitation to a passive rotational joint which makes the shape of the elastic strip just before snap-through buckling complicated. The range limitation leads to storing the higher elastic energy for an impulsive motion without changing the actuator of the active joint. We show the results of theoretical analysis and numerical simulations which support the effectiveness of the proposed robotic catapults. We also show that the jumping robot with the asymmetric robotic catapult achieves high performance.
This paper proposes a novel walking gait generation for a double-pendulum-like biped walking model based on the family of the symmetric orbits. By introducing an involution R associated with the leg switching map, the flow of the saddle-center dynamics of the uncontrolled pendulum possesses the time-reversal symmetry with respect to R, and the conjunction of the flow and R forms a family of symmetric orbits parameterized by the orbital energy and the stride bound. The invariant manifold of the family of symmetric orbits is obtained numerically or approximately using a perturbation method. By constraining the solution onto the invariant manifold using the control inputs, stable walking gaits are generated in a semi-global manner. Based on the passivity of the closed-loop system, a robust speed-controlled walking is achieved in a very simple way.
This paper presents an approach for developing a mechanical system that is similar to a musclo-skeletal system. Antagonistic driving of articulation is a key mechanism, in which at least two muscles control a single rotary axis in the antagonistic manner. It provides not only to control the joint angle but also to control the joint stiffness. The stiffness control requires for individual muscle to have a non-linear elasticity. Hence we have developed an actuator system having non-linear elasticity, which is called ANLES. It fundamentally mimics the skeletal muscle in the sense of having non-linear elasticity. This paper describes a method for designing the elastic characteristics of ANLES and shows the developed ANLES followed by the theoretical and experimental result of its non-linear elasticity. Next one rotary joint controlled by two ANLESes is introduced. The experimental result of measuring stiffness of the joint is shown. A method for controlling the joint angle is proposed, in which two ANLESes drives one rotary joint with no angular feedback. An on-line method for estimating the weight loaded at the end-point is also proposed. All of methods are evaluated and validated by the experiments.