Generally, real time scheduling algorithms provide optimal solution when the application context is well known and its behavior completely mastered. But in practice, both the real time constraints and the tasks' characteristics are often imprecisely known. This results in sub-optimal behaviors generating failures that should be avoided. In this paper, a natural approach based on fuzzy calculus is proposed and compared with two others: the flexible/fuzzy constraints satisfaction problems, and probability based model. It can be shown that the fuzzy calculus is as efficient as the others, but simpler. This approach allows more realistic knowledge representations and definitions of more flexible real time scheduling algorithms. >