The next generation of ground-based optical/infrared (IR) telescopes will have primary mirrors of up to 42 m. To take advantage of the large potential increase in angular resolution, adaptive optics will be essential to overcome the resolution limits set by atmospheric turbulence. Novel techniques such as Multi-Conjugate Adaptive Optics (MCAO) and Multi-Object Adaptive Optics (MOAO) are being developed to achieve near diffraction-limited images over large fields-of-view. This thesis concerns the development of MCAO and MOAO pathfinders. Specifically, the construction of CANARY, aMOAO demonstrator, and the on-sky performance and scientific exploitation of the Multi-conjugate Adaptive optics Demonstrator (MAD). CANARY is under construction for the William Herschel Telescope (WHT) in La Palma and contains a telescope simulator to allow testing of the set-up in the laboratory. The simulator contains a natural guide star emulator, turbulence phase screens, and telescope relay optics. The work presented here concerns the integration of the various components in relation to numerical models and the CANARY specifications. MAD was a near-IR imager on the Very Large Telescope (VLT) in Chile. Science demonstration observations were taken of R136, the young, massive cluster situated in the 30 Doradus star-forming region in the Large Magellanic Cloud. These data were used here to determine the MCAO performance across the ~1’x1’ field-of-view, for different pointings with respect to the guide stars, finding high Strehl ratios and relatively uniform corrections across the fields. The MAD data are then used to construct radial surface brightness profiles for R136, providing new insights into intriguing past results from the Hubble Space Telescope. The MAD data reveal that the profile is strongly asymmetric, removing the need for dramatic dynamical evolution of the cluster in the recent past, and highlighting the importance of considering asymmetries when analysing clusters further afield. The MAD data, combined with other near-IR imaging from the VLT, are then used to investigate the nature of candidate young stellar objects from recent observations with the Spitzer Space Telescope.