Only the downlink in LTE supports multi-antenna processing in the form of spatial multiplexing for a single UE. So far, two multi-antenna related techniques are explicitly supported in the LTE standard for uplink:
UE transmit antenna switching
In antenna switching, the UE may transmit over either of two transmit antennas. This corresponds to a poor man’s version of transmit diversity since it makes it possible, at least in principle, to avoid implementing two separate RF chains and instead use an RF switch to select which antenna should be used. The scheme is supposed to be closed-loop in the sense that the eNodeB can decide at sub-frame speed which of the two possible antennas should be used, presumably based on some measurements on uplink sounding reference symbols.
MU-MIMO for uplink
Similar to MU-MIMO in the downlink, in principle nothing prevents scheduling multiple UEs on the same resource blocks also in the uplink. Different UEs often have considerably different channels and this can be exploited to co-schedule UEs with orthogonal channels for increased system capacity.
Interference rejection techniques in the eNodeB can be used to reduce the requirements on the orthogonally of the channels. Thus, MU-MIMO in the uplink is just a fancy term for well-known SDMA using interference rejection.
In contrast to MU-MIMO for the downlink, the entire receiver processing is performed at a single location, which is at the eNodeB. The eNodeB has a complete information about the transport formats used by all the UEs and knows the channel corresponding to each UE. This is a distinct advantage that improves the range of applicability of MU-MIMO in the uplink. Nevertheless, MU-MIMO by nature targets high load scenarios so an important issue is to determine at which system load does uplink MU-MIMO becomes beneficial.