This study presents a theoretical method for planning and controlling agile bipedal locomotion based on robustly tracking a set of non-periodic keyframe states. Based on centroidal momentum dynamics, we formulate a hybrid phase-space planning and control method that includes the following key components: (i) a step transition solver that enables dynamically tracking non-periodic keyframe states over various types of terrain; (ii) a robust hybrid automaton to effectively formulate planning and control algorithms; (iii) a steering direction model to control the robot's heading; (iv) a phase-space metric to measure distance to the planned locomotion manifolds; and (v) a hybrid control method based on the previous distance metric to produce robust dynamic locomotion under external disturbances. Compared with other locomotion methodologies, we have a large focus on non-periodic gait generation and robustness metrics to deal with disturbances. This focus enables the proposed control method to track non-periodic keyframe states robustly over various challenging terrains and under external disturbances, as illustrated through several simulations.
An accelerated learning algorithm (ABP-adaptive back propagation) is proposed for the supervised training of multilayer perceptron networks. The learning algorithm is inspired from the principle of "forced dynamics" for the total error functional. The algorithm updates the weights in the direction of steepest descent, but with a learning rate a specific function of the error and of the error gradient norm. This specific form of this function is chosen such as to accelerate convergence. Furthermore, ABP introduces no additional "tuning" parameters found in variants of the backpropagation algorithm. Simulation results indicate a superior convergence speed for analog problems only, as compared to other competing methods, as well as reduced sensitivity to algorithm step size parameter variations.
This paper introduces research leading to a computer-aided design tool in which engineering designers can test various design concepts (topologies) in an environment equipped to automatically model the dynamics and conveniently optimize the specified components (given the evaluation criteria defined by human designers). A component repository is developed to store not only the component dynamics models, but also other information such as typical component design constraints and physical constitutive laws. In this paper, automated modeling of design configurations is introduced through a design representation called a conceptual dynamics graph (CD graph) and generic models of various components. CD graphs contain the information on how physical components as well as their generic models are topologically connected. A generic component model can accommodate various types of coupling between this component and its environment. This paper also discusses a systematic approach to automatically prepare a mechatronic design problem for the use of optimization to tune the parameters for optimum dynamics. Since genetic algorithms are used for this optimization, this preparation decodes and encodes proper design variables into design genotypes while taking into consideration the design constraints and physical constitutive laws.
Experimental results are described in which a rod of magnetostrictive terfenol was used in the dual capacity of a passive structural support element and an active vibration control actuator and artificial neural networks were used for the adaptive real-time control algorithm. Tests were performed on a three-legged table, where the terfenol actuators mentioned above are the table legs. For the table experiment, shaker vibrations generated in the ground and transmitted to the tabletop (via the legs) were attenuated by counter vibrations synthesized in the table leg actuators. The goal of this experiment was to maintain a quiescent tabletop in the presence of floor vibrations.Utilizing a proportional-integral derivative and a neural network controller, actuated forces were used to cancel applied disturbance forces. The neural network architecture identifies (learns) and adapts to the tabletop forced disturbance through a fast adaptation law known as Adaptive BackPropagation, generating the required counter vibration. The architecture and hence the control was designed to be modular so cross talk (coupling in the control signal) is minimized. This puts an extra burden on the controller to decouple the spillovers but maintains modularity, an important feature for large scale implementations.This article describes work in this area and demonstrates the ability to cancel disturbances from static to the hundred hertz frequency range.
Abstract:Sliding mode control is examined from the perspective of obtaining stable and robust tracking of an arbitrary time-varying reference by a multi-input-output, linear, time-invariant system driven by a certain class of bounded errors, nonlinearities and disturbances. Most existing schemes for such systems are subsumed by the one presented here. The results are developed via the use of inverse models, and make clear the constraints imposed by the finite and infinite zero structure of the system. In particular, stable and robust tracking is shown to be obtained by the scheme in this paper if and only if the system is minimum phase.
scite is a Brooklyn-based organization that helps researchers better discover and understand research articles through Smart Citations–citations that display the context of the citation and describe whether the article provides supporting or contrasting evidence. scite is used by students and researchers from around the world and is funded in part by the National Science Foundation and the National Institute on Drug Abuse of the National Institutes of Health.