Bioinspired Neurodynamics Based Intelligent Tracking Control Of Autonomous Surface Vehicles

Abstract:Autonomous surface vehicles (ASVs) are kind of small surface vessels, which can operate autonomously on the water and complete various complex tasks. Because of their outstanding characteristics in small volume, flexible mobility, high hiding power and unmanned operation, ASVs have great perspective and application in civil utilization, military and national defense construction. However, the control problem of ASVs is still very challenging, as it involves the topics of marine systems, robotics and nonlinear control.Trajectory tracking control is one of important research areas in the control of ASVs. It requires the vehicles to reach and follow the desired time-parameterized trajectory within limited time until arriving at the desired position. Since ASVs are nonlinear systems strongly coupling with the water, it is very difficult to obtain an accurate model description of ASVs, especially the hydrodynamic parameters that must be obtained by repeated experiments. Furthermore, most of ASVs in real application are kind of underactuated mechanical systems, which makes the tracking control of ASVs even more challenging.Motivated by the above considerations, this thesis takes an in-depth study on the tracking control of ASVs. Bioinspired neurodynamics based intelligent approaches are proposed for fully actuated and underactuated ASVs with unknown dynamics. The proposed approaches take advantage of the unique continuous and smooth features of the neural dynamics model derived from membrane equation for a biological neural system.The main results and innovations are included as follows.1) a bioinspired neurodynamics based approach is proposed for tracking control of ASVs subject to unknown ocean currents. For ASVs subject to unknown ocean currents, where smooth and continuous velocity commands are desirable for safe and effective operation, a novel biologically inspired tracking control approach is proposed. First, the tracking controller is designed from the error dynamics of the ASVs by constructing a Lyapunov function that stablizes the tracking errors. Then three shunting neural dynamics models are incorporated as components in the controller. A simple observer is designed to estimate the unknown ocean currents. The stability of the overall control system under the controller and observer is guaranteed by a Lyapunov theory. The developed tracking controller is capable of generating smooth and continuous control signals even in the presence of unknown ocean currents.2) an efficient neural network (NN) approach is proposed for tracking control of ASVs with unknown dynamics.For fully actuated ASVs with unknown vehicle dynamics and subject to significant uncertainties, an efficient NN tracking controller is proposed. The proposed NN has a single-layer structure by utilizing the vehicle regressor dynamics that expresses the highly nonlinear dynamics in terms of the known and unknown dynamic parameters. The learning algorithm of the NN is derived from Lyapunov stability analysis, which is simple yet computationally efficient. A robust compensator is also designed for the external disturbances. The proposed NN approach tracks the desired trajectory with good control performance through the on-line learning of the NN without any off-line learning procedures. In addition, the proposed controller is capable of compensating bounded unknown disturbances.3) a novel backstepping neurodynamics based approach is proposed for tracking control of underactuated ASVs.For underactuated ASVs, backstepping is a very effective control methodology, but the fundamental drawback is that it requires to calculate numerical derivatives of virtual velocity control signals. To overcome this problem, a bioinspired neurodynamics based approach is proposed by integrating the backstepping technique and three shunting neural dynamics models. Instead of differentiating the virtual control signal directly at each step of the backstepping design procedure, we let the corresponding virtual signal pass through a shunting neuron model to avoid the complexity of differentiation. As a result, the proposed tracking control algorithm is much simpler than that constructed based on the conventional backstepping approach.4) a biologically inspired approach is proposed for tracking control of underactuated ASVs with unknown dynamics.For underactuated ASVs with unknown dynamics, a novel biologically inspired tracking control approach is proposed by integrating backstepping technique, shunting neural dynamics model and NN. The shunting neural dynamics models are incorporated to solve the inherent complexity problem of numerical derivatives of virtual control signals in the backstepping design, and the NN is employed for the unknown dynamic parameters. Compared to other methods, the proposed tracking control approach is computationally efficient as no derivative calculations on virtual controls are required. In addition, it tracks the reference trajectory without any prior knowledge of the dynamics parameters.