Ch. 16 — Ad Hoc Wireless Networks

Chapter 16 of Wireless Communications (2nd ed. Draft) — Andrea Goldsmith

Abstract

This chapter establishes the fundamental capacity limits and energy-efficient design principles for ad hoc wireless networks, focusing on the impact of network size and mobility. It presents the seminal scaling laws showing that per-node throughput decays asymptotically with the number of nodes unless mobility is exploited. Furthermore, it details energy-constrained optimization strategies, specifically analyzing the trade-off between transmit power and circuit processing energy across protocol layers. These results are critical for designing battery-limited wireless sensor networks and mobile ad hoc systems where energy conservation dictates network lifetime.

Key Concepts

• Capacity Region Scaling Laws: For a network of $K$ nodes, the per-node throughput scales as $1/\sqrt{K\log K}$ to $1/\sqrt{K}$ as $K\to \infty$. This result indicates that without mobility, individual node rates approach zero despite optimal routing, as resources are consumed by forwarding traffic for others. The total network throughput grows only as $\sqrt{K}$, highlighting the congestion challenges in large dense networks.
• Mobility-Induced Capacity: Node mobility can increase the per-node rate to a constant value by exploiting variations in network topology. This improvement follows because mobility allows nodes to physically transport information closer to destinations, reducing multi-hop forwarding overhead. However, exploiting these variations incurs significant delay, establishing a fundamental throughput-delay trade-off.
• Achievable Rate Regions: Capacity is often characterized by computing rate regions based on suboptimal transmission strategies like time-division. A rate matrix describes the set of sustainable rates for all source-destination pairs at a given moment, where the achievable region is the convex combination of these matrices over time. Strategies including power control, spatial reuse, and successive interference cancellation expand the set of usable rate matrices.
• Energy-Constrained Circuit Power: In short-range applications, circuit energy consumption for signal processing is comparable to transmit power, necessitating joint optimization. Minimizing transmission time to utilize sleep modes often yields greater energy savings than minimizing transmit power alone. This shifts design focus from pure transmit efficiency to total energy consumption per bit.
• Cooperative MIMO: Nodes can cooperate to form virtual multiple-input multiple-output (MIMO) systems even if individual nodes possess single antennas. This requires exchanging information between cluster nodes, which is energy-efficient only if the inter-node distance is small relative to the transmission distance. Energy savings are achieved when the transmit/receive clusters are separated by 10 to 20 times the distance between cooperating nodes.
• Cross-Layer Energy Optimization: Energy efficiency requires optimization across all protocol layers because trade-offs between energy, delay, and throughput are interdependent. For instance, reducing transmit energy by slowing transmission rate may increase congestion delay at the network layer. Application requirements dictate the optimal operating point on the trade-off curve.

Key Equations and Algorithms

• Simplified Path-Loss Model: $P_r=P_t{d}^{-\gamma }$. This equation models signal propagation where received power $P_r$ is determined by transmit power $P_t$, distance $d$, and path-loss exponent $\gamma$. It serves as the foundational physical layer model for calculating interference and required transmit power in the capacity and routing analyses.
• Per-Node Throughput Scaling: $R_{node}\propto 1/\sqrt{K}$. This relationship quantifies the Gupta and Kumar result for asymptotic network capacity in random ad hoc networks. It demonstrates that as the number of nodes $K$ increases, the maximum sustainable data rate for any single node decreases.
• Energy per Bit Calculation: $E_b=P/C$. This expression relates received energy per bit $E_b$ to received power $P$ and data rate $C$. It is used to derive the minimum energy required for reliable communication in the wideband limit by taking the limit as bandwidth approaches infinity.
• SINR Power Control Matrix: $(I-F)P\ge u$. In power control problems, this matrix inequality represents the signal-to-interference-plus-noise ratio (SINR) constraints for multiple users. $I$ is the identity matrix, $F$ is the interference matrix, $P$ is the power vector, and $u$ represents the target SINR requirements normalized by noise.
• Minimum Energy per Bit (Wideband Limit): The minimum energy per bit required for reliable communication in the limit of infinite bandwidth is constant regardless of fading if the receiver lacks channel knowledge. Specifically, $E_b/N_0$ converges to a value determined by the noise spectral density, achievable via on-off signaling.
• Cooperative MIMO Distance Condition: $d_{cluster}\ll 0.1\cdot d_{dist}$. Energy efficiency in cooperative MIMO requires the distance between cooperating nodes $d_{cluster}$ to be significantly smaller than the transmission distance $d_{dist}$. If this condition is not met, the energy cost of local exchange exceeds the benefits of multiplexing or diversity.
• Routing Cost Function: Cost = $w_1\cdot E+w_2\cdot D$. Routing protocols minimize a weighted sum of energy consumption $E$ and delay $D$. The weighting factors $w_1$ and $w_2$ allow the network to trade off energy efficiency against latency based on application constraints such as real-time video or sensor monitoring.
• Capacity per Unit Energy Definition: Maximum bits transmitted per unit energy such that error probability $\to 0$ as energy $\to \infty$. Unlike Shannon capacity, this metric applies to hard energy-constrained systems where finite energy prevents asymptotic zero-error transmission. It prioritizes spectral efficiency at ultra-low power regimes.

Key Claims and Findings

• In large fixed ad hoc networks, the per-node data rate approaches zero as the number of nodes increases, regardless of optimal routing strategies. This degradation is caused by intermediate nodes dedicating most of their resources to forwarding packets for other nodes rather than their own data.
• Mobility can increase the per-node rate to a constant value, but this capacity gain is exchanged for increased packet delay. The throughput-delay trade-off characterizes the limit of performance as node mobility introduces beneficial topological variations.
• For short-range battery-limited devices, circuit energy consumption is often comparable to transmit energy, making total energy minimization more critical than transmit power minimization alone. Techniques like M-ary modulation reduce transmission time, allowing circuitry to sleep sooner and save energy.
• On-off signaling is near-optimal for minimizing energy per bit in the wideband limit, achieving capacity per unit energy even with unknown fading at the receiver. While coherent modulation is optimal with channel knowledge, simple signaling suffices when bandwidth is the dominant resource.
• In the wideband limit, time-division multiple access (TDMA) is the optimal strategy for minimizing energy per bit for broadcast and multiple access channels. However, in finite bandwidth regimes, superposition strategies like CDMA may achieve the same energy efficiency with less bandwidth.
• Cooperative MIMO provides energy savings only when the overhead of local information exchange is small compared to the transmission distance savings. If cooperating nodes are too far apart relative to the transmission range, the energy cost of coordination negates the MIMO gains.

Terminology

• Capacity Region: The set of all maximum data rates possible between all source-destination pairs in a network. For $K$ nodes, this region has dimension $K(K-1)$ for independent information or is larger if common information (multicasting) is sent.
• Rate Matrix: A matrix describing the set of rates that can be sustained simultaneously by all source-destination pairs at any specific moment in time. It functions as a fundamental unit for computing achievable rate regions via time-division convex combinations.
• Capacity per Unit Energy: A reliability metric defined as the maximum number of bits per unit energy that can be transmitted with vanishing error probability as energy grows asymptotically large. It is the relevant capacity definition for systems with hard energy constraints rather than power constraints.
• Spatial Reuse: A capacity enhancement technique where simultaneous transmissions occur in different parts of the network provided they do not interfere destructively. It allows the network to utilize time and frequency resources more efficiently by overlapping non-conflicting links.
• Circuit Energy Consumption: The power consumed by analog and digital signal processing circuitry during packet transmission and reception, separate from transmit power. This component becomes dominant in short-range low-power systems, influencing modulation and sleep mode design.
• Wideband Limit: The theoretical operational regime where channel bandwidth $B$ approaches infinity while data rate $C$ approaches zero. In this limit, the minimum energy per bit is independent of fading statistics and achievable with simple signaling schemes.
• Cooperative MIMO: A technique where geographically distributed single-antenna nodes coordinate to act as a virtual multi-antenna array. This allows the system to exploit MIMO gains without requiring hardware complexity at individual nodes.
• Energy-Constrained Routing: Routing protocols that optimize paths based on minimizing energy consumption per hop or maximizing the minimum battery lifetime across all nodes. Cost functions may combine energy and delay to balance network longevity with quality-of-service requirements.