Ee Power Control Eval

ee-power-control-eval

Energy-Efficient Power Control: A Look at 5G Wireless Technologies — Zappone et al. (2015) (arXiv:1503.04609, 2015)

What this evaluates

This evaluation probes the energy efficiency and feasibility of power control algorithms in 5G massive MIMO and relay-assisted interference networks. It measures how well centralized and distributed algorithms maximize Global Energy Efficiency (GEE) while satisfying minimum per-user rate constraints under hardware impairments and Rayleigh fading.

Datasets

• Hardware-Impaired Massive MIMO System — total ?; splits: test (-1)
• Relay-assisted OFDMA interference network — total ?; splits: test (-1)

Metrics

• Average GEE (primary) — range: bit/s/Hz/Watt
  • Global Energy Efficiency (GEE) is defined as the ratio of the total achievable data rate to the total power consumption (including transmit and circuit power). It is averaged over 50,000 independent channel and user-drop scenarios.
• Probability of feasibility — range: [0, 1]
  • The fraction of independent scenarios (over 50,000 trials) where the EE maximization problem satisfies the minimum per-user rate constraints.
• Convergence iterations — range: integer
  • Average number of algorithm iterations required until the relative change in power allocation vector satisfies $| \mathbf{q}^{(i)} - \mathbf{q}^{(i-1)}|^2 / | \mathbf{q}^{(i)}|^2 \leq 10^{-4}$.

Input / output format

Input: Channel coefficients (Rayleigh fading), hardware impairment magnitudes ($\epsilon_{BS}$), maximum transmit power ($\overline{P}$), circuit power ($p_{c,k}$), and minimum rate constraints ($R_k%$).

Output: Optimal power allocation vector $\mathbf{p}$ (or $\mathbf{q}$), along with computed GEE, minimum rate, and feasibility status.

Scoring recipe

def compute_metrics(power_alloc, channel_params, impairment, max_power, circuit_power, min_rate_pct):
    rates = compute_spectral_efficiency(power_alloc, channel_params, impairment)
    total_rate = sum(rates)
    total_power = sum(power_alloc) + len(rates) * circuit_power
    gee = total_rate / total_power
    max_possible_rate = compute_rate_at_infinity(channel_params, impairment)
    min_rate = min(rates)
    feasible = min_rate >= (min_rate_pct / 100) * max_possible_rate
    return gee, feasible, min_rate

Common pitfalls

• When QoS constraints are infeasible, the algorithm falls back to the unconstrained ($R=0%$) solution, which can artificially inflate reported GEE in low-power regimes.
• Distributed algorithms do not jointly manage interference, leading to significant GEE degradation at high transmit powers compared to centralized schemes.
• Convergence is declared based on a relative change threshold of $10^{-4}$, which may mask slow convergence or oscillation in dense interference scenarios.

Evidence (verbatim from paper)

The results are obtained by averaging over $5 \cdot 10^4$ independent scenarios of users' drops and channel coefficients. As seen, $\mathcal{P}_f$ approaches 1 for realistic values of $\overline{P}$ up to $25%$ of the maximum rate... Fig. 2 shows the average GEE for the following resource allocation policies... Convergence is declared when $| \mathbf{q}^{(i)} - \mathbf{q}^{(i-1)}|^2 / | \mathbf{q}^{(i)}|^2 \leq 10^{-4}$.

Citation

@misc{zappone2015energy,
  title={Energy-Efficient Power Control: A Look at 5G Wireless Technologies},
  author={Zappone et al. (2015)},
  year={2015},
  note={arXiv:1503.04609}
}

• arXiv: 1503.04609