International audience

International audience

International audience

The main objective of this paper is to present a new accurate power profiler for embedded systems and smartphones. The second objective is, for it, to be a tutorial explaining the main steps to build power profilers for embedded and mobile systems, in general. We start our work by firstly describing the general methodology of building a power profiler. Then, we showcase how each step is undertaken to build a profiler with two power models. The first one was an artificial neural network (called N2) that presented a lot of noise in its estimation. After debugging and improvement, the second model, a NARX neural network (we call N3) was built. It eliminated all the drawback of the first model and had a mean absolute percentage error of 2.8%.

International audience

International audience

14th International Conference on Informatics in Control, Automation and Robotics (ICINCO)

This paper focuses on a control approach for legged robots with fast gaits. The method consists in a sliding mode based control extended the consideration of system energy. The control objective is then to perform stable trajectories or controlled limit cycles (CLC), by means of energy regulation with respect to a shaped nominal representation (implicitly related to system trajectories). In this paper, we establish some theoretical results. Then, some capabilities of the approach are illustrated through examples related to spring-mass systems usually considered to characterize some legged robots behaviors.

In some mechanical systems which make very small motions like industrial wagons or constructor robots, the dispositive of their control usually presents some imperfections, a succession of jumps and stops, when the static friction force is larger than the dynamic friction force. This effect is called the stick-slip phenomenon. In this paper, we develop a nonlinear observer in order to estimate the friction force of the contact during the motion, and to compensate the effects which cause the stick-slip phenomenon.

A neural network adaptive force controller is proposed for a real hydraulic system. The dynamic model of this system is highly non-linear and very complex to obtain. Thus, it is considered as a black box, and a priori identification becomes necessary. A neural network is used to approximate the model, then a controller using the Lyapunov approach is designed. The neural network parameters are updated online according to an adaptation algorithm obtained via stability analysis. The performance of the proposed neural network controller is validated on an experimental plant.

Control of legged robots with fast gaits is addressed in this paper. These kind of systems interact intermittently with environment. We develop a new approach for control of legged robots by use of controlled limit cycles (CLC). The designed control system generates the desired trajectories (online) and control input. Viable definition of gaits and their admissibility are introduced.

This paper deals with the application of the H/sub /spl infin//-approach to robustly control the force exerted by a hydraulic servo-actuator on its environment. In order to evaluate the system stability and the performance, large environmental uncertainties and disturbances are introduced by use of a particular workpiece. The major nonlinearities of the system are pointed out through experimental tests and parameters estimation results. Then, a linear time-invariant model for control is defined with respect to both system dynamics and application requirements. Experimental and simulation control results are presented and the achievement of a stability/performance trade-off is discussed.

Describes a control/command architecture developed for the authors' quadruped robot RALPHY (French acronym for legged robot with pneumatic actuators). Dynamical and real time aspects as well as theoretical and practical approaches have been taken into account. In order to hierarchize and distribute as far as possible the different tasks to be realized, the robot dynamic model has been decomposed in two parts. A "coordinator level" controls the platform stability by means of leg to platform interaction model. In a dynamic situation, this stability is ensured by a "center of mass acceleration control" and a "force distribution control". A "leg level" deals with the control problem of each leg. The authors propose an adaptive computed torque type controller which takes into account nonlinearities due both to the dynamic model and to the pneumatic system. Finally, the authors develop the connection between these two levels, proposing a solution to deal with uneven terrains.

In this paper we propose an adaptive impedance control scheme consisting of two stages. The first stage performs an online estimation of the robot parameters imposing the desired mechanical impedance. It constitutes the kernel of the control system and remains active during the complete, free or constrained, motion of the robot. A simple algorithm for the numerical computation of the defined impedance error is presented where only the available feedback information is used. The imprecision in the parameters of the environment (stiffness, positioning) is compensated by the second stage of adaptation. This one constitutes an external force control loop closed around the internal nonlinear impedance controller. Simulation results obtained for a single leg of a pneumatic driven, quadruped robot show the effectiveness of the proposed control scheme in case of considerable uncertainty both in the robot and ground parameters.

Makes a comparative study between two adaptive control strategies applied to a leg of a quadruped. This leg is composed of two rigid links with pneumatic actuators driving rotational joints. In order to enable nonlinear control approaches, a model of this system is built. The control design is based on one hand on MRAS and on the other hand on passive system theory which ensure the asymptotic stability of the whole. Adaptation laws are proposed, grounded either on the number of operations keeping in mind real time implementation facilities, or on the robustness toward disturbances. These control laws are compared using simulation results.>

The behavior of autoregressive moving average (ARMA) predictors is analyzed for neglected dynamic cases with nonstationary signals. For this analysis, nonstationarity types must be considered (abrupt or smooth changes in the mean and variance). Some procedures can be used to improve the algorithm performances and limit the self-stabilization oscillations. The resulting algorithm is not much more complex than the NLMS algorithm and can be used for applications like ADPCM or data compression with highly nonstationary signals.>

Adaptive comb filters (ACFs) have notches at multiples of the fundamental frequency. For the ACF proposed only one parameter is adapted for removal of harmonic components. The RML (recursive maximum-likelihood) algorithm used is much less complex than the existing ones. The ACF is implemented as cascaded second-order cells and only the parameter of the first cell is estimated. Depending on the energy of the fundamental frequency, convergence to a local minimum can be realized. Some strategy (penalty) can be implemented to avoid this problem. It can be shown that the proposed algorithm is globally stable provided that the debiasing parameter is appropriately chosen. No stability test is needed. If p is the number of harmonic components, only 12p multiplications and 10p additions are required for the RML estimation of ACF parameters.>

The authors consider adaptive prediction with an IIR (infinite impulse response) moving average (MA) part, controlled either by the recursive least-mean squares (LMS) algorithm or by an extended least squares (ELS) algorithm based on a posteriori errors. The predictor input is either the sum of bandpass components or a nonstationary speech sentence. The bounded input/bounded output (BIBO) stability is a major issue. Although both algorithms enjoy the self-stabilization property when the input is stationary, it is shown that with a nonstationary input the LMS algorithm can be unstable due to power jumps in the speech. The ELS algorithm always ensures BIBO stability.>

To provide frequency estimation, adaptive notch filters implemented in a cascade of second-order cells can be used, along with strategies to track time-varying parameters. The estimated frequencies are then used to provide the estimation of amplitudes and phases in a recursive manner. The proposed RML (recursive maximum-likelihood) algorithm consists of two steps. The first step involves a maximum-likelihood algorithm to adapt the cascaded filter parameters which will provide the frequency estimates. The second step uses these estimates, and the amplitudes and phases are estimated by a recursive-least-squares algorithm The RML algorithm is asymptotically consistent and robust with respect to the neglected dynamics. In the case of time-varying signals its tracking capabilities ensure the goodness of the estimations.>

This paper presents several structures for Adaptive Notch Filters and implementation aspects of constrained ARMA models. These ANF suit very well to evolutionary spectrum analysis for sinusoidal signals in additive noise. colored or white. Several estimation algorithms may be applied with these structures (ELS,RML,RIV) and some special procedures may be imployed to increase the convergence rate, enhance the estimation consistency and reduce tha algorithm complexity. Lastly, non-stationary ANF are considered and an estimation procedure implemented.