In Multiple-Input Multiple-Output (MIMO) radar, adaptive multi-target tracking can be achieved using subspace tracking algorithms in conjunction with super-resolution subspace localisation algorithms. However, the presence of unknown radar clutter deteriorates or breaks algorithms that rely on the white noise assumption. In this paper, an H ∞ approach is proposed for robust tracking of each target's range, Direction of Arrival (DOA) and velocity with unknown clutter. Specifically, Doppler SpatioTemporal ARray (Doppler-STAR) and virtual Doppler-STAR manifold extenders are first proposed by combining the slow-time and fast-time dimensions of a pulse MIMO radar received signal. Then, in the 'extended' space, H ∞ adaptive filtering is used to track a noise subspace equivalent, which exists regardless of the noise assumption. Finally, the target parameters are extracted from the adaptively tracked noise subspace. Based on computer simulation studies, the performance of the proposed H ∞ tracking approach is evaluated using challenging tracking scenarios and compared against several existing subspace tracking algorithms.