TL;DR
This paper introduces Serial-Max, a low-complexity antenna selection algorithm for Full Duplex (FD) MIMO systems. By selecting bidirectional communication links in a two-step "greedy" fashion rather than through exhaustive search, the authors achieve near-optimal data rates and error performance while drastically reducing the computational burden on the hardware.
What Problem Does This Paper Address?
In traditional wireless systems, we use Half Duplex (HD): you either talk or listen, but not both at the same time on the same frequency. Full Duplex (FD) is the "holy grail" of spectral efficiency, allowing simultaneous bidirectional communication. However, FD faces a massive hurdle: Self-Interference (SI). Imagine trying to hear a friend's whisper while you are shouting at the top of your lungs—that is what a receiver deals with in an FD system.
While modern suppression techniques can dampen this "shouting," some residual interference always remains. In a MIMO system (multiple antennas at both ends), we can improve performance by selecting the best specific antennas for transmitting and receiving. The "optimal" way to do this is an exhaustive search—checking every possible combination of antenna pairs. As the number of antennas grows, this becomes a "combinatorial explosion" that is too slow and power-hungry for real-time mobile devices.
Methods
The authors propose the Serial-Max selection algorithm. Instead of looking at all possible combinations of bidirectional links simultaneously, it breaks the problem into two simple steps:
- Step 1: Scan all possible links between Node A and Node B and pick the single best one (the one with the highest "obtainable SINR").
- Step 2: Remove the antennas used in Step 1 from the pool. Then, pick the best remaining link from the remaining sub-matrix.
To handle different priority levels for the two directions (e.g., if the download needs to be faster than the upload), they incorporate a time-sharing factor.
Crucially, the authors don't just propose the algorithm; they provide a rigorous mathematical framework. They derive closed-form expressions for the average weighted sum rate and the Symbol Error Rate (SER) using order statistics and the law of total probability. This allows engineers to predict system performance without running millions of time-consuming simulations.
Datasets
This is a theoretical and simulation-based study. No real-world datasets were used. Instead, the authors evaluated their method using:
- Monte-Carlo Simulations: Used to validate the derived mathematical formulas.
- Channel Models: They assumed independent Rayleigh fading for the communication links and modeled the residual self-interference as a Rayleigh-distributed random variable, which is standard practice for effectively suppressed FD systems.
Key Insights and Ideas
- The "Greedy" Advantage: The most surprising insight is Lemma 1, which proves that the Serial-Max method is actually optimal if the second-best link in the system is either the second or third largest element in the original channel matrix.
- Asymptotic Optimality: The authors show that as you add more antennas to the system, the probability that Serial-Max picks the "wrong" pair drops quadratically. In other words, for large MIMO arrays, the simple greedy approach is virtually indistinguishable from the perfect brute-force search.
- The Rate Ceiling: The paper highlights that FD systems are fundamentally limited by a "rate ceiling." Unlike HD systems, where increasing power usually increases the rate, in FD, increasing power also increases the self-interference. Eventually, the rate flattens out, making antenna selection even more critical to maximize that ceiling.
Key Results
- Complexity Reduction: For a system with $N$ antennas at each node, the exhaustive search requires roughly $O(N^4)$ comparisons. Serial-Max reduces this to $O(N^2)$. For a 12x12 antenna system, this is the difference between ~10,000 operations and ~250 (a 40x speedup).
- Performance Gap: Simulations show that the Serial-Max method achieves almost the same weighted sum rate as the optimal Max-WSR (Weighted Sum Rate) method across all SNR regions.
- Diversity Order: With perfect interference cancellation, the system achieves a diversity order of $(N_A-1)(N_B-1)$. However, the authors note that if residual interference exists, the diversity order technically drops to zero because of the "error floor" caused by the interference.
Limitations and Future Work
- Single RF Chain Constraint: The model assumes each node has only one transmit and one receive RF chain. In modern "Massive MIMO," we might want to select multiple transmit/receive pairs simultaneously, which would require a more complex version of this algorithm.
- Perfect Channel Estimation: The analysis assumes the system knows the channel state perfectly at the start of each time slot. In reality, channel estimation error is a significant factor in FD systems.
- Static Weights: The weights for bidirectional traffic are assumed to be given. Future work could look at dynamic weight adjustment based on real-time buffer status or varying user requirements.
Summary updated