In reply to Swetha_ch:
Achieving 100% coverage is the “holy grail” of verification. Achieving 100% coverage is not only hard, but confidence that you achieved 100% coverage means you are done is an even harder challenge.
Running the same tests multiple times with different seed values is the simplest approach that might increase your coverage. How effective that is depends on your design and the way you have architected the testbench. You need to look at the range random values produced your constraint to make sure they are as random as you expected them to be.
There is a lot of research going into machine-learning techniques to improve coverage by looking at which random variables have the most impact on coverage results, but there is nothing I would call production worthy yet. I would say we are still in the human learning stage of machine learning—trying different algorithms and seeing which ones lead to better results.