Recent experiments on the ASDEX Upgrade tokamak aim at improving the physics base for ITER and DEMO to aid the machine design and prepare efficient operation. Type I edge localized mode (ELM) mitigation using resonant magnetic perturbations (RMPs) has been shown at low pedestal collisionality . In contrast to the previous high ν* regime, suppression only occurs in a narrow RMP spectral window, indicating a resonant process, and a concomitant confinement drop is observed due to a reduction of pedestal top density and electron temperature. Strong evidence is found for the ion heat flux to be the decisive element for the L–H power threshold. A physics based scaling of the density at which the minimum PLH occurs indicates that ITER could take advantage of it to initiate H-mode at lower density than that of the final Q = 10 operational point. Core density fluctuation measurements resolved in radius and wave number show that an increase of R/LTe introduced by off-axis electron cyclotron resonance heating (ECRH) mainly increases the large scale fluctuations. The radial variation of the fluctuation level is in agreement with simulations using the GENE code. Fast particles are shown to undergo classical slowing down in the absence of large scale magnetohydrodynamic (MHD) events and for low heating power, but show signs of anomalous radial redistribution at large heating power, consistent with a broadened off-axis neutral beam current drive current profile under these conditions. Neoclassical tearing mode (NTM) suppression experiments using electron cyclotron current drive (ECCD) with feedback controlled deposition have allowed to test several control strategies for ITER, including automated control of (3,2) and (2,1) NTMs during a single discharge. Disruption mitigation studies using massive gas injection (MGI) can show an increased fuelling efficiency with high field side injection, but a saturation of the fuelling efficiency is observed at high injected mass as needed for runaway electron suppression. Large locked modes can significantly decrease the fuelling efficiency and increase the asymmetry of radiated power during MGI mitigation. Concerning power exhaust, the partially detached ITER divertor scenario has been demonstrated at Psep/R = 10 MW m−1 in ASDEX Upgrade, with a peak time averaged target load around 5 MW m−2, well consistent with the component limits for ITER. Developing this towards DEMO, full detachment was achieved at Psep/R = 7 MW m−1 and stationary discharges with core radiation fraction of the order of DEMO requirements (70% instead of the 30% needed for ITER) were demonstrated. Finally, it remains difficult to establish the standard ITER Q = 10 scenario at low q95 = 3 in the all-tungsten (all-W) ASDEX Upgrade due to the observed poor confinement at low βN. This is mainly due to a degraded pedestal performance and hence investigations at shifting the operational point to higher βN by lowering the current have been started. At higher q95, pedestal performance can be recovered by seeding N2 as well as CD4, which is interpreted as improved pedestal stability due to the decrease of bootstrap current with increasing Zeff. Concerning advanced scenarios, the upgrade of ECRH power has allowed experiments with central ctr-ECCD to modify the q-profile in improved H-mode scenarios, showing an increase in confinement at still good MHD stability with flat elevated q-profiles at values between 1.5 and 2.