What do you mean by "online adapt"? Do you want to adapt the parameters dynamically? However, a possible starting point for your research can be the following source: http://iridia.ulb.ac.be/irace/
Your question is an important and interesting one:
online parameter control may be useful to better adapt an algorithm to the characteristics of a particular instance-this may be particularly beneficial if the instances are relatively heterogeneous, requiring different settings of key algorithm parameters to reach peak performance. Additionally, it may be desirable to adapt algorithm parameters depending on the stage of the search process, for example, to make a transition from a rather explorative to an exploitative search phase, or depending on the amount of infeasibility when dealing with (strictly) constrained problems.
You may look at the useful link below:
Article Parameter Control in Evolutionary Algorithms: Trends and Challenges