Process changes, such as flow disturbances and sensor noise, are common in the chemical and metallurgical industries. To maintain optimal performance, the controlling system has to adapt continuously to these changes. This is a difficult problem because the controller also has to perform well while it is adapting. The Adaptive Neural Swarming (ANS) method introduced in this paper satisfies these goals. Using an existing neural network controller as a starting point, ANS modifies the network weights through Particle Swarm Optimisation. The ANS method was tested in a real-world task of controlling a simulated non-linear bioreactor. ANS was able to adapt to process changes while simultaneously avoiding hard operating constraints. This way, ANS balances the need to adapt with the need to preserve generalisation, and constitutes a general tool for adapting neural network controllers on-line.

PDF, PS

In Proceedings of the Genetic and Evolutionary Computation Conference, William B. Langdon and Erick Cantu-Paz and Keith E. Mathias and Rajkumar Roy and David Davis and Riccardo Poli and Karthik Balakrishnan and Vasant Honavar and G{"u}nter Rudolph and Joachim Wegener and Larry Bull and Mitchell A. Potter and Alan C. Sch (Eds.), pp. 13 2002.

@InProceedings{conradie:gecco02, title={Adaptive Control Utilising Neural Swarming}, author={Alex v. E. Conradie and Risto Miikkulainen and Christiaan Aldrich}, booktitle={Proceedings of the Genetic and Evolutionary Computation Conference}, editor={William B. Langdon and Erick Cantu-Paz and Keith E. Mathias and Rajkumar Roy and David Davis and Riccardo Poli and Karthik Balakrishnan and Vasant Honavar and G{"u}nter Rudolph and Joachim Wegener and Larry Bull and Mitchell A. Potter and Alan C. Sch}, pages={13}, url="http://www.cs.utexas.edu/users/ai-lab?conradie:gecco02", year={2002} }