This paper introduces dotRL, a platform that enables fast implementation and testing of Reinforcement Learning algorithms against diverse environments. dotRL has been written under .NET framework and its main characte...
详细信息
This paper introduces dotRL, a platform that enables fast implementation and testing of Reinforcement Learning algorithms against diverse environments. dotRL has been written under .NET framework and its main characteristics include: (i) adding a new learning algorithm or environment to the platform only requires implementing a simple interface, from then on it is ready to be coupled with other environments and algorithms, (ii) a set of tools is included that aid running and reporting experiments, (iii) a set of benchmark environments is included, with as demanding as Octopus-Arm and Half-Cheetah, (iv) the platform is available for instantaneous download, compilation, and execution, without libraries from different sources.
Repetitive processes propagate information in two separate directions, one of which is temporal and the other can be spatial. A repetitive process makes a series of sweeps, or passes, through a set of dynamics over th...
详细信息
This paper is dedicated to the asymptotic stability of 2D discrete Roesser models. Two well-known necesssary and sufficient conditions expressed in terms of characteristic polymials are recalled and their equivalence ...
This paper is dedicated to the asymptotic stability of 2D discrete Roesser models. Two well-known necesssary and sufficient conditions expressed in terms of characteristic polymials are recalled and their equivalence is proved. Although it is not new result, the provided proof is simpler than those proposed in the literature. These conditions being numerically rather non tractable, a first motivation is to compare these conditions through their sufficient LMI (Linear Matrix Inequalities) relaxations. Actually, the same approach, based upon the S-procedure, is used to derive the two relaxations, which, once again, are proved to be equivalent. However, the second condition offers a different point of view which leads to a modification of the relaxation technique, making the necessity of the LMI condition reachable.
An important part of the modern computer networks design is to develop novel technologies, architectures and control mechanisms for network devices enabling power saving by adapting network capacities to current traff...
详细信息
An important part of the modern computer networks design is to develop novel technologies, architectures and control mechanisms for network devices enabling power saving by adapting network capacities to current traffic loads and user demands. We describe centralize and hierarchical control frameworks for reducing power consumption in backbone computer networks. The implementation of these frameworks provides the local control mechanisms that are implemented in the network devices level and network-wide control strategies implemented in the central control level. In this paper, we focus on network-wide algorithms for calculating the power status of network devices and the energy-aware MPLS routing for recommended network configuration. We enumerate several possible formulations of a network energy saving optimization problem with continuous and discrete variables. We discuss the limitations of these approaches and problems with their application to power control in real networks. We propose the relaxation of the complete binary problem formulation assuming full routing and energy state of all devices calculation, and the algorithm to solve it. Our formulation is based on a heuristic approach that leads to a continuous optimization. The evaluation of the optimization scheme through simulation is presented in the final part of the paper.
The paper presents a new approach to identification of Wiener systems based on instrumental variables method. It is assumed that the linear dynamic system is represented by the discrete transfer function and the inver...
详细信息
The paper presents a new approach to identification of Wiener systems based on instrumental variables method. It is assumed that the linear dynamic system is represented by the discrete transfer function and the inverse characteristic of the nonlinear element is represented by any set of specified basis functions. It is shown that parameters of a modified series-parallel Wiener model estimated using the least squares method are inconsistent. To obtain consistent parameter estimates, the instrumental variables method is employed. The instrumental variables are generated by passing the system input through the linear dynamic model obtained with the least squares method. It is also shown that proposed identification method can be extended to Wiener systems with inverse nonlinear characteristics that does not contain the first order term. A simulation example is also included to show the effectiveness and practical feasibility and illustrate asymptotic convergence properties of the proposed approach.
Prototyping Systems of Concurrent Cyclic Processes (SCCP), in which a number of simultaneously running processes compete to access to a set of resources, in an automated way is the main thread of this paper. Procedure...
详细信息
Prototyping Systems of Concurrent Cyclic Processes (SCCP), in which a number of simultaneously running processes compete to access to a set of resources, in an automated way is the main thread of this paper. Procedure of performance evaluation for such systems with desired values of a wide range of system functional characteristics, such as a schedule of processes flows and the system period, especially with multiple executions of single process cycles within single system cycle will be presented. This stage is realised by a procedure of automatic building of an analytical model of SCCP, which is based on the (max, +) algebra formalism. In most cases the (max, +) algebra is used to model of discrete-event systems, while the synchronization is based on randez-vous protocol. However, in this paper - modeling of the systems is based on mutual exclusion protocol, moreover buffers are included. Moreover the algorithm of determining of the system period and schedule of workflow - when there is no direct relation between eigenvalue of system matrix and system period - will be presented.
The paper concerns the task of robust fault diagnosis of actuators in non-linear discrete-time systems. The general unknown input observer strategy and the H ∞ framework are utilised to design a robust fault detecti...
详细信息
The paper concerns the task of robust fault diagnosis of actuators in non-linear discrete-time systems. The general unknown input observer strategy and the H ∞ framework are utilised to design a robust fault detection scheme. Moreover, the proposed approach enables to perform the isolation of the faulty actuators. The final part of the paper presents an illustrative example which exhibits the performance of the proposed approach.
Recently, the theory of Wave Repetitive Processes has been applied to derive a control for a class of ladder circuits that features wave-like active inter-stage connections [1]. There, the discretization of the circui...
Recently, the theory of Wave Repetitive Processes has been applied to derive a control for a class of ladder circuits that features wave-like active inter-stage connections [1]. There, the discretization of the circuit has been implemented by means of a basic difference scheme. Wave Digital Filter (WDF) theory provides a more elaborate scheme that ensures a number of favourable properties, and is typically used for strictly passive circuits. In this paper, the adaptability of such an active circuit by means of Multidimensional Wave Digital Filters (MDWDF) is analysed with regard to the implementation of non-standard elements as multiple sources and intra-structural secondary connections. Conclusively, the simulation results are studied and compared.
The paper deals with the problem of robust fault estimation of non-linear discrete-time systems. In particular, it is shown how to employ the unknown input observer approach and the H ∞ strategy to design a robust f...
详细信息
ISBN:
(纸本)9781479901890
The paper deals with the problem of robust fault estimation of non-linear discrete-time systems. In particular, it is shown how to employ the unknown input observer approach and the H ∞ strategy to design a robust fault estimation filter. The proposed approach is designed in such a way that a prescribed disturbance attenuation level is achieved with respect to the fault estimation error while guaranteeing the convergence of the observer. The resulting design procedure is relatively simple and boils down to solving a set of linear matrix inequalities, which can be efficiently achieved with modern computational packages. The final part of the paper presents an illustrative example which exhibits the performance of the proposed approach.
Repetitive processes make a series of sweeps, or passes, through a set of dynamics defined over the finite pass length. Once a pass is complete the process resets to the starting location ready for the start of the ne...
Repetitive processes make a series of sweeps, or passes, through a set of dynamics defined over the finite pass length. Once a pass is complete the process resets to the starting location ready for the start of the next one. On any pass, the output, or pass profile, generated on the previous one acts as a forcing function on, and therefore contributes to, the dynamics of the new pass profile. These processes are a class of 2D systems where the novel control problem is that the sequence of pass profiles generated can contain oscillations that increase in amplitude from pass-to-pass. These oscillations cannot be removed by standard control action and therefore a 2D systems control theory for them has been under development over the pass two decades. This paper demonstrates that, unlike much of the control theory for other 2D systems, repetitive processes have applications that have led to experimental verification of control laws.
暂无评论