Policy iteration for optimal switching with continuous-time dynamics

Author(s):  
Tohid Sardarmehni ◽  
Ali Heydari
Author(s):  
Tohid Sardarmehni ◽  
Ali Heydari

Two approximate solutions for optimal control of switched systems with autonomous subsystems and continuous-time dynamics are developed. The proposed solutions consist of online training algorithms with recursive least squares training laws. The first solution is the classic policy iteration algorithm which imposes heavy computational burden (full back-up). In order to relax the computational burden in the policy iteration algorithm, the second algorithm is presented. The convergence of the proposed algorithms to the optimal solution in online training is investigated. Simulation results are presented to illustrate the effectiveness of the discussed algorithms.


The major goal of this paper is to explore the effective state estimation algorithm for continuous time dynamic system under the lossy environment without increasing the complexity of hardware realization. Though the existing methods of state estimation of continuous time system provides effective estimation with data loss, the real time hardware realization is difficult due to the complexity and multiple processing. Kalman Filter and Particle Filer are fundamental algorithms for state estimation of any linear and non-linear system respectively, but both have its limitation. The approach adopted here, detect the expected state value and covariance, existed by random input at each stage and filtered the noisy measurement and replace it with predicted modified value for the effective state estimation. To demonstrate the performance of the results, the continuous time dynamics of position of the Aerial Vehicle is used with proposed algorithm under the lossy measurements scenario and compared with standard Kalman filter and smoothed filter. The results show that the proposed method can effectively estimate the position of Aerial Vehicle compared to standard Kalman and smoothed filter under the non-reliable sensor measurements with less hardware realization complexity.


Complexity ◽  
2021 ◽  
Vol 2021 ◽  
pp. 1-24
Author(s):  
Héctor A. Echavarria-Heras ◽  
Cecilia Leal-Ramírez ◽  
Guillermo Gómez ◽  
Elia Montiel-Arzate

We examine the comportment of the global trajectory of a piecewisely conceived single species population growth model. Formulation relies on what we develop as the principle of limiting factors for population growth, adapted from the law of the minimum of Liebig and the law of the tolerance of Shelford. The ensuing paradigm sets natality and mortality rates to express through extreme values of population growth determining factor. Dynamics through time occur over different growth phases. Transition points are interpreted as thresholds of viability, starvation, and intraspecific competition. In this delivery, we focus on the qualitative study of the global trajectory expressed on continuous time and on exploring the feasibility of analytical results against data on populations growing under experimental or natural conditions. All study cases sustained fittings of high reproducibility both at empirical and interpretative slants. Possible phase configurations include regimes with multiple stable equilibria, sigmoidal growth, extinction, or stationarity. Here, we also outline that the associating discrete-time piecewise model composes the logistic map applied over a particular region of the phase configuration. Preliminary exploratory analysis suggests that the logistic map’s chaos onset could surpass once the orbit enters a contiguous phase region.


2008 ◽  
Vol 58 ◽  
pp. 143-152
Author(s):  
Paolo Arena ◽  
Davide Lombardo ◽  
Luca Patanè

In this contribution a survey on a novel approach to locomotion and perception in biologically inspired robots is presented. The basic electronic architecture for modeling and implementing nonlinear dynamics involved in motion and perceptual control of the robot is the Cellular nonlinear network paradigm. It is shown how this continuous time lattice of neural-like circuits can generate suitable and real-time dynamics for efficient control of multi-actuators moving machines, and also to create the basis for a perceptual control of their behaviors.


Sign in / Sign up

Export Citation Format

Share Document