Tuning the seismic control systems in order to achieve optimal performance isa challenging area due to the system and disturbance uncertainties. Although,model uncertainties, process time delay, and actuator dynamics can be consid-ered as typical uncertainties, the main source of uncertainty in a seismic con-trol system comes from the aleatory nature of earthquake disturbances. Inthis case, tuning of the control system based on a given seismic record maynot necessarily result in optimal performance for other earthquakes. In thispaper, a methodological approach is proposed for online control of active struc-tural control systems considering seismic uncertainties. For this purpose, theconcept of reinforcement learning is utilized for online tuning of an activemass drive system. The controller comprises a gain‐scheduling fuzzy propor-tional derivative controller whose gains are tuned via an online reinforcementlearning algorithm. Moreover, in order to tackle the time delays, a dynamicstate predictor is utilized in conjunction with the proposed controller. To eval-uate the performance of proposed controller, according to an assumed site haz-ard, thousand ground motion records are generated and clustered based ontheir spectral features using a fuzzy clustering approach. Finally, the controlleris implemented in a laboratory‐scale structure, and its performance is exam-ined in the presence of the cluster centers and some real seismic records sim-ulated on a shake table. The test results reveal successful performance of theproposed controller in tackling a wide range of seismic disturbances in thepresence of time delay
