(Partial) Unit Memory ((P)UM) codes are a special rate class of convolutional codes with memory m=1. (P)UM codes allow to construct convolutional codes based on block codes, which makes and algebraic description and efficient decoding possible. There exist several constructions based on Reed--Solomon or BCH codes.

Gabidulin codes are a special class of so-called rank metric codes. Recently, they have attracted attention since they provide an almost optimal solution for error control in random linear network coding. In the context of network coding, dependencies between different blocks transmitted over a network can be created by defining rank metric convolutional codes.

Recently, metrics for convolutional codes in rank metric were defined and two constructions based on Gabidulin codes were presented. Moreover, an extension of a decoding algorithm (the Dettmar--Sorger algorithm) to rank metric was shown.

In this thesis, the concept of Unit Memory codes is extended to Multi Memory codes. First, a construction of (Partial) Multi Memory codes based on Gabidulin codes is presented. Then, the distance parameters of this construction are estimated based on the known rank metric definitions for convolutional codes. Second, the Dettmar--Sorger algorithm is adapted to rank metric based (Partial) Multi Memory Codes of arbitrary rate. The correctness of the decoding algorithm is proven and it is shown that its complexity is cubic with the length.