The paper studies the motion control for mobile relays imple-
menting cooperative beamforming to aid the communication
between a source-destination pair. We consider an urban com-
munication scenario, where the channels exhibit spatiotempo-
ral correlations and thus can be learned. The relays move in a
time-slotted fashion within a three-dimensional cube. During
every slot, the relays beamform optimally to maximize the
Signal-to-Interference+Noise Ratio (SINR) at the destination
and decide their positions for the next slot. Unlike prior works
that assume knowledge of channel statistics, our proposed ap-
proach is model-free. Also, typically, prior approaches as-
sume discrete motion on the two-dimensional plane. How-
ever, as discretization introduces the curse of dimensionality,
those methods do not easily extend to three-dimensional mo-
tion. We propose a model-free, continuous control actor-critic
approach that can be easily applied to 2D and 3D motion with
the same complexity . To address the random nature of the
channel, we propose to use Sinusoidal Representation Net-
works (SIRENs) for value function approximation. Our ap-
proach outperforms the direct application of the State-of-the-
Art continuous control algorithms for both 2D and 3D cases.