1,721,104 research outputs found
Value iteration via output feedback for LQ optimal control of SISO systems
No full textIn this paper, a value iteration algorithm, which makes use of just input/output measurements, is proposed to solve linear quadratic (LQ) optimal control problems for single-input, single-output (SISO) plants. This algorithm is designed by coupling an adaptive Luenberger observer with an indirect value iteration architecture for continuous-time plants. The effectiveness of the proposed approach is validated via numerical simulations
Patrolling and collision avoidance beyond classical Navigation Functions
No full textIn this paper, we consider the problem of navigating a single unicycle-like robot, while avoiding obstacles in a known environment, and, at the same time, of steering the agent itself to monitor and patrol an assigned path. To this end, we propose a novel framework that combines tools and algorithms borrowed from algebraic geometry with techniques inspired by those associated with classical navigation functions. The former aspect permits the systematic construction of Lyapunov functions that certify the convergence with an assignable decaying rate to the desired patrolling path in the absence of obstacles. This control action is then combined with an additional term and a supervisory logic obtained by relying on the collision avoiding abilities of the underlying navigation function. Such a mixed strategy may potentially lead beyond the current understanding and implementation of classical navigation functions. The paper is then concluded by several numerical simulations that corroborate the theoretical results
Value iteration for linear quadratic optimal control of single-input single-output systems via output feedback
Full text linkA value iteration approach based solely on input/output measurements is proposed to solve linear quadratic (LQ) optimal control problems for single-input, single-output (SISO) continuous-time systems. Such an algorithm is designed by coupling an adaptive Luenberger observer with an indirect value iteration architecture. The continuous-time implementation of this controller requires that the gathered estimates are strongly controllable. A hybrid adaptation mechanism is envisioned to overcome such a requirement. The effectiveness of the proposed approach is validated via numerical simulations
The Price of Information in dynamic games on networks
No full textThe main objective of this paper consists in providing definitions and preliminary results that allow to assess and evaluate the importance of information and data exchange patterns in non-cooperative dynamic games defined over (communication or physical) networks. This objective is achieved by introducing suitable metrics that associate a value to each communication link among the agents. In the case of Linear-Quadratic (LQ) games on networks, necessary and sufficient conditions are then provided that allow to characterize the set of all Nash equilibria that can be generated for a given network topology, namely even in the presence of partial and limited information
Motion Planning, Formation Control and Obstacle Avoidance for Multi-Agent Systems
No full textIn this paper we investigate the problem of navigating, in a centralized manner, a team of unicycle-like robots in a known environment avoiding obstacles and, at the same time, steering them to patrol an assigned path in controlled formation. Such a goal is pursued by combining algorithms that use tools borrowed from algebraic geometry with some methods inspired by classical navigation functions. The former allows to automatically construct Lyapunov functions that certify the convergence to the desired formation and to the desired path, if obstacles are absent, while the latter allow to avoid collisions among agents and with fixed obstacles. Simulations are reported throughout all the paper to illustrate the theoretical results
Deterministic Optimality of the Steady-State Behavior of the Kalman-Bucy Filter
Full text linkIn this letter, we provide a deterministic characterization of optimality of the steady-state behavior of the Kalman-Bucy filter, via an inverse optimal control argument. The result is achieved in two steps, both interesting per se. First, a singular linear-quadratic (LQ) optimal control problem is formulated and solved with respect to the innovation term of a classic Luenberger observer, hence yielding a LQ optimal observer. Then, such a construction is employed to interpret the optimality of the steady-state behavior of the celebrated Kalman-Bucy filter in a purely deterministic sense
On the use of difference of log-sum-exp neural networks to solve data-driven model predictive control tracking problems
No full textWe employ Difference of Log-Sum-Exp neural networks to generate a data-driven feedback controller based on Model Predictive Control (MPC) to track a given reference trajectory. By using this class of networks to approximate the MPC-related cost function subject to the given system dynamics and input constraint, we avoid two of the main bottlenecks of classical MPC: the availability of an accurate model for the system being controlled, and the computational cost of solving the MPC-induced optimization problem. The former is tackled by exploiting the universal approximation capabilities of this class of networks. The latter is alleviated by making use of the difference-of-convex-functions structure of these networks. Furthermore, we show that the system driven by the MPC-neural structure is practically stable
Q-learning for continuous-time linear systems: a data-driven implementation of the Kleinman algorithm
No full textA data-driven strategy to estimate the optimal feedback and the value function in an infinite-horizon, continuous-time, linear-quadratic optimal control problem for an unknown system is proposed. The method permits the construction of the optimal policy without any knowledge of the model, without requiring that the time derivatives of the state are available for the design, and without even assuming that an initial stabilizing feedback policy is available. Two alternative architectures are discussed: the first scheme revolves around the periodic computation of some matrix inversions involving the Q-function, whereas the second approach relies on a purely continuous-time implementation of some dynamic systems whose trajectories are uniformly attracted by the solutions to the above algebraic equations. Interestingly, the proposed strategy essentially constitutes a (direct) data-driven implementation of the celebrated Kleinman algorithm, hence subsuming the particularly appealing features of the latter, such as quadratic monotone convergence to the optimal solution. The theory is then validated by the means of practically motivated applications
Adaptive high-gain observers for systems in observability canonical form
No full textThis paper deals with the problem of designing observers for nonlinear systems in observability canonical without having at one's disposal a model for the underlying dynamics. Such an objective is pursued by coupling the classical highgain observer structure with several different adaptation laws borrowed from the adaptive control literature. The overall architecture is demonstrated to accomplish the set goal under rather standard assumptions. Some numerical simulations are provided to demonstrate the effectiveness of the proposed observer
- …
