Résumé
After a successful development of theoretical and numerical works on Fermi acceleration at relativistic shocks, some difficulties recently raised with the scattering issue, a crucial aspect of the process. Most pioneering works were developed assuming the scattering off magnetic fluctuations as given. Even in that case, when a mean field is considered, its orientation is mostly perpendicular to the shock normal in the front frame, and this tends to quench the scattering process. Solving this difficulty leads to address the issue of the generation of very intense magnetic fluctuations at short wave lengths. The relativistic motion of the shock front let the cosmic rays to visit upstream during a very short time only, making this generation of magnetic fluctuations very challenging. Anyway there is some hope to solve the problem. Thanks to a recent work by Spitkovsky (2008) \cite{AS}, we know that the process works without any mean field and now we have to investigate up to which intensity the mean field can be amplified for allowing Fermi process with appropriate fast instabilities. In this presentation, the collisionless shock structure in relativistic regime is sketched, the scattering issue is presented, and the instabilities that can provide the expected magnetic field amplification are presented as well. Although there exists observational evidence that particles are accelerated in relativistic flows and are distributed according to a power law suggesting a Fermi process, the drastic conditions for Fermi process to work are not always clearly fulfilled.