Nowadays, optimization is begun to be use in different fields, e.g. preference algorithms. These new challenges need a robustness meta heuristics to solve them. Supernova meta heuristic that emules the descent behavior of the gradients and share the same weakness of them. They get stuck planar regions and hardly find the needle minimum. The main objective of this works is to improve the performance of the original version of supernova for the problematic topologies mention above. First, a review of how to these problems are solved in the literature is presented. Second, A criterion to determine planar regions is described . Third, a strategy to choose the parameters agree with the topology of the function is implemented. Supernova 2.0 was tested using the set of benchmarks functions proposed in CEC2013. The new version is significantly better than the original version, no significantly better than SPSO2011 and significantly inferior with SADE. Although, the results are applied to Supernova, most of the strategies can be applied to other methods.