Nowadays, robotic systems are more and more complex. The complexity in a robotic system may come from several sources, including (i) a complex structure (high number of degrees of freedom, flexibility, singularities, friction effects, complex design, small/big workspace, etc), (ii) its dynamics (highly nonlinear dynamic model, coupled dynamics, unknown and/or time-varying parameters, etc), (iii) its actuation (under-actuation or over-actuation, complex actuation, actuation with unilateral constraints, control input saturations, etc), (iv) interaction with its environment (internal and external disturbances, unknown/unusual environment, time-varying operating conditions, etc), and (v) real-time constraints (i.e. computing time, sample time, etc). These systems are gaining more and more attention from both academic and industrial communities. However, their control problems cannot be considered as a simple task due to the above-mentioned challenges, that should be taken into account during the control design stage to ensure the desired performances. Within this context, my research has focused on the development of new nonlinear control schemes (mainly adaptive, robust or predictive) and their real-time application to complex robotic systems. These activities can be split up into five areas of research in robotics, namely (i) Control of rigid parallel kinematic manipulators (PKMs), (ii) Control of cable-driven parallel robots (CDPRs), (iii) Control of tethered autonomous underwater vehicles, (iv) Control of fin-actuated bio-inspired underwater vehicles, and (v) Control of rehabilitation wearable exoskeletons. All my proposed control contributions are validated through real-time experiments in different operating conditions.