ABSTRACT
In this chapter, we shall first explore the utilization of adaptive NNs within the event-triggered feedback controller framework to achieve precise trajectory tracking in two robotic applications. In the first application, we consider the robotic manipulators and detail the design procedure as presented in Narayanan et al., 2018. Specifically, we shall explore the output feedback control scheme that incorporates a nonlinear NN observer to reconstruct the joint velocities of the manipulator using joint position measurements. In addition to the observer NN, a second NN is employed to compensate for nonlinearities in the robot dynamics. We will consider two distinct configurations for the control scheme, depending on whether the observer is co-located with the sensor or the controller within the feedback loop. For both configurations, the controller computes the torque input by leveraging the observer NN and the second NN. We shall derive the eventtriggering condition and weight update rules for the controller and observer using the Lyapunov stability method.
In the second application, we shall see the application of event-sampled output-feedback NNbased controller for steering a quadrotor Unmanned Aerial Vehicle (UAV). The controller design encompasses multiple components to ensure precise flight control. Firstly, an observer is devised to estimate the UAV’s state-vector from its outputs. Subsequently, a kinematic controller is developed to determine the desired translational velocity, which is used in conjunction with a virtual controller to establish the desired rotational velocity for the UAV’s orientation convergence. The signals generated by the dynamic controller enable the tracking of the desired lift velocity and rotational velocities. Throughout the designs, the impact of sampling errors is emphasized. This part of the chapter is based on the results presented in Szanto et al., 2017a,b.
Finally, we shall explore the application of event-triggered ADP-based controllers in cyberphysical systems (CPS) wherein multiple real-time dynamic systems are connected to their corresponding controllers through a shared communication network. In particular, we shall look at a distributed scheduling protocol design via cross-layer approach to optimize the performance of CPS by maximizing the utility function that is generated by using the information collected from both the application and network layers. We shall see that when compared with traditional scheduling algorithms, the application of event-triggered control synthesis together with a distributed scheduling scheme via the cross-layer approach presented here can not only allocate the network resources efficiently but also improve the performance of the overall real-time dynamic system. This application presented in this chapter is based on the work in Xu, 2012 and Xu and Jagannathan, 2012.
