Adaptive model-based neural network control: validation and analysis

An adaptive model-based neural network controller (AMBNNC) uses multilayer perceptron artificial neural networks to enhance the high-speed trajectory-tracking accuracy of robotic manipulators. The artificial neural networks are trained through repetitive training on trajectory-tracking error data to provide an estimate of payload. The payload estimate adapts the feedforward compensator to unmodeled system dynamics and payload variation. The result is a computationally efficient direct form of adaptive control. The AMBNNC concept was previously validated for a single joint. Here, experimentation and analysis are extended to the first three links of a PUMA-560 manipulator. Two forms of neural network payload estimation are investigated. Tracking performance is evaluated for a wide range of payload and trajectory conditions and compared with that of a nonadaptive model-based controller. The performance improvement potential and the limitations of the AMBNNC approach are illustrated