This paper explores the use of neural networks to solve highly nonlinear control problems, particularly in the context of self-learning control systems. The authors demonstrate how a multilayered neural network can learn to control a nonlinear dynamic system by first emulating the system's dynamics and then using the emulator to train a controller. The controller is another multilayered neural network that learns to control the emulator, which in turn controls the actual dynamic system. The learning process continues as the emulator and controller improve and track the physical process. The paper introduces the Adaline, a processing element with variable weights, and the back-propagation algorithm for training layered neural networks. The Adaline's output is a weighted sum of its inputs, passed through a sigmoid function, which allows it to behave linearly for small inputs and saturate for larger inputs. The back-propagation algorithm updates the weights to minimize the mean-square error between the desired and actual outputs. The control problem addressed is the steering of a trailer truck while backing up to a loading dock. The truck emulator is trained to predict the next state of the truck given the current state and steering input. The neural network controller is then trained to drive the emulator from an initial state to a desired state using the emulator as a guide. The training process involves multiple runs, each starting from a random initial state, and the controller's weights are updated using back-propagation to minimize the error. An example of a truck backer-upper is provided, where the controller successfully steers the truck to the dock from various initial positions. The controller's performance is evaluated using a root-mean-square error metric, showing that it can achieve high accuracy in controlling the truck's position and orientation. The paper concludes by discussing the potential of this approach for solving a wide range of nonlinear control problems and outlines future research directions, including the determination of the complexity of the emulator relative to the system being controlled.This paper explores the use of neural networks to solve highly nonlinear control problems, particularly in the context of self-learning control systems. The authors demonstrate how a multilayered neural network can learn to control a nonlinear dynamic system by first emulating the system's dynamics and then using the emulator to train a controller. The controller is another multilayered neural network that learns to control the emulator, which in turn controls the actual dynamic system. The learning process continues as the emulator and controller improve and track the physical process. The paper introduces the Adaline, a processing element with variable weights, and the back-propagation algorithm for training layered neural networks. The Adaline's output is a weighted sum of its inputs, passed through a sigmoid function, which allows it to behave linearly for small inputs and saturate for larger inputs. The back-propagation algorithm updates the weights to minimize the mean-square error between the desired and actual outputs. The control problem addressed is the steering of a trailer truck while backing up to a loading dock. The truck emulator is trained to predict the next state of the truck given the current state and steering input. The neural network controller is then trained to drive the emulator from an initial state to a desired state using the emulator as a guide. The training process involves multiple runs, each starting from a random initial state, and the controller's weights are updated using back-propagation to minimize the error. An example of a truck backer-upper is provided, where the controller successfully steers the truck to the dock from various initial positions. The controller's performance is evaluated using a root-mean-square error metric, showing that it can achieve high accuracy in controlling the truck's position and orientation. The paper concludes by discussing the potential of this approach for solving a wide range of nonlinear control problems and outlines future research directions, including the determination of the complexity of the emulator relative to the system being controlled.