An intelligent controller for neurophysiological experiments

The authors describe an intelligent controller (ICON) for neurophysiology experiments that employs approximate reasoning techniques and a dynamic control planner (DYNCON) to perform real-time analysis and control of the experiment. Results from experiments simulated from real data show that ICON reached the same conclusions as did the investigator during and after the actual experiment but that ICON did so in fewer experimental trials. An evaluation showed that ICON´s performance in controlling the experiment was limited by the time required to collect the experimental trial data (experiment time) and not the time required to analyze the data (analysis time); analysis time was always less than 11% of the experiment time, indicating that the current hardware and software technology is fast enough for real-time control of experiments similar to the one described. The evaluation also showed that the time spent in the DYNCON was from 4.5% to 15% of the analysis time. The results indicate that advantages achieved by the DYNCON, such as greater flexibility, responsivity to incoming data, and adaptability to the changing demands placed on the system as the experiment progresses, outweigh the cost in terms of computation time