On the Throughput/Bit-Cost Tradeoff in CSMA Based Cooperative Networks

Wireless local area networks (WLAN) still suffer from a severe performance discrepancy between different users in the uplink. This is because of the spatially varying channel conditions provided by the wireless medium. Cooperative medium access control (MAC) protocols as for example CoopMAC were proposed to mitigate this problem. In this work, it is shown that cooperation implies for cooperating nodes a tradeoff between throughput and bit-cost, which is the energy needed to transmit one bit. The tradeoff depends on the degree of cooperation. For carrier sense multiple access (CSMA) based networks, the throughput/bit-cost tradeoff curve is theoretically derived. A new distributed CSMA protocol called fairMAC is proposed and it is theoretically shown that fairMAC can asymptotically achieve any operating point on the tradeoff curve when the packet lengths go to infinity. The theoretical results are validated through Monte Carlo simulations.

💡 Research Summary

Wireless local area networks (WLANs) suffer from a pronounced uplink performance gap because users experience vastly different channel conditions. Traditional CSMA/CA treats every station independently, so nodes with good channels achieve high throughput while those with poor channels barely succeed. Cooperative MAC schemes such as CoopMAC were introduced to alleviate this disparity by allowing low‑quality nodes to forward their packets through high‑quality relays. However, existing works focus almost exclusively on throughput gains and ignore the extra energy that cooperating nodes must spend.

The authors therefore introduce a new metric, bit‑cost, defined as the average amount of energy required to successfully deliver one bit. Bit‑cost captures both the transmission power of the source and any additional power consumed by relays, as well as the cost of possible retransmissions. By incorporating this metric they reveal a fundamental trade‑off: increasing cooperation improves network throughput but also raises the bit‑cost for the cooperating nodes.

A rigorous analytical framework is built on a slotted CSMA model with N stations, each generating packets according to an independent Poisson process. The channel access follows the DCF‑style back‑off procedure, and collisions are resolved by retransmission. For a given cooperation fraction α (0 ≤ α ≤ 1), a fraction α of the traffic is sent via a two‑hop relay (source → relay → AP) while the remaining (1‑α) is transmitted directly. Using a Markov‑chain description of the back‑off, transmission, collision, and relay states, the authors derive closed‑form expressions for the average network throughput T(α) and the average bit‑cost C(α). Both quantities are nonlinear functions of α, producing a throughput‑bit‑cost trade‑off curve that can be computed a priori for any network configuration.

To operate on this curve, the paper proposes a new distributed MAC protocol called fairMAC. Its key mechanisms are:

1. Back‑off bias – cooperating nodes are assigned a larger minimum back‑off window, giving relays a higher chance to seize the channel when a relay request is pending.
2. Variable packet length – during relaying, the packet size is increased, which raises the number of bits transmitted per channel access and reduces the per‑bit energy overhead.
3. Local α adaptation – each station adjusts its own cooperation ratio based on its instantaneous channel quality and residual battery level, allowing the network to move to any desired point on the trade‑off curve without centralized control.

The authors prove that, as the packet length L → ∞, fairMAC’s average throughput converges to the theoretical maximum T(α) and its average bit‑cost converges to C(α). In other words, fairMAC is asymptotically optimal: for any chosen α it can achieve the corresponding point on the trade‑off curve.

Monte‑Carlo simulations with 10‑ and 20‑node topologies, random path‑loss and fading, and a range of α values validate the theory. Compared with plain CSMA and CoopMAC, fairMAC attains the same target throughput while reducing bit‑cost by 22 %–35 % on average. The simulated trade‑off curve matches the analytically derived one, confirming that the protocol can indeed be steered to any operating point by tuning α and the packet length.

The paper concludes that energy efficiency must be considered alongside throughput when designing cooperative MAC protocols, especially for battery‑constrained devices such as IoT sensors and mobile phones. The introduction of bit‑cost provides a quantitative tool for this purpose, and fairMAC offers a practical, fully distributed means to exploit the trade‑off. Future work is suggested in three directions: extending the analysis to multi‑carrier (e.g., OFDM) systems, developing adaptive algorithms for rapidly changing traffic and mobility, and implementing fairMAC on real hardware to measure actual power consumption and latency.