Temel Kavramlar
The author proposes a new approach to compute the majority vote (MV) function based on modulation on conjugate-reciprocal zeros (MOCZ) without requiring channel state information at the transmitters and receiver.
Özet
The author introduces three methods to compute the MV function in an over-the-air computation (OAC) scenario:
Uncoded MV Computation (Method 1):
The transmitters map their votes to the zeros of Huffman polynomials.
The receiver evaluates the polynomial constructed by the superposed sequence at the conjugate-reciprocal zero pairs and detects the MV using a direct zero-testing (DiZeT) decoder.
This method provides the highest computation rate but requires the power-delay profile (PDP) information at the receiver.
Differential MV Computation (Method 2):
The transmitters use a differential encoding to map their votes to the zeros.
The receiver uses the DiZeT decoder to detect the MVs without needing the PDP information.
This method reduces the computation rate by half compared to Method 1.
Index-based MV Computation (Method 3):
The transmitters map their votes to the zeros based on an index calculated using all the votes.
The receiver uses the DiZeT decoder to detect the MVs without needing the PDP information.
This method provides better computation error rate (CER) performance than Methods 1 and 2 by exploiting redundancy, but at the expense of a lower computation rate.
The author theoretically analyzes the CERs of the proposed methods and demonstrates their efficacy in a distributed median computation scenario in a fading channel. The proposed methods are robust to phase and time synchronization errors as they do not require channel state information at the transmitters and receiver.