In this text, we have observed that cooperation enables non-trivial network performance benefits. But one may wonder if cooperative communications will be implemented in mainstream networking technologies or if cooperation will remain just a research curiosity. From the perspective of practice, a contrarian can identify many potential barriers to implementation. We examine two such barriers.
Channel Variation Speed
Coherent combining will never work because distributed phase synchronization is too difficult.
This statement is likely valid in mobile environments, since phase synchronization requires tracking movements to within a fraction of a wavelength. However, for fixed wireless systems with a dominant line-of-sight path, this issue needs further investigation. Another potentially feasible scenario would be mixed wireline/wireless networks in which multiple base stations are connected by a fiber-optic backbone that enables sufficient information exchange for coherent transmission.
Cooperative diversity mechanisms are useful only if a node is unable to exploit the temporal variation of the channel. In typical mobile environments, the time variation is more than sufficient for a node to achieve diversity gains using coding and/or ARQ without employing a cooperative relay.
The main issue is still the speed of channel variations. If the channel varies sufficiently quickly, the value of cooperation seems to diminish. However, for applications requiring small delay, the value of cooperative diversity is easily recognized.
Processing Energy
The energy consumption in decoding overheard packets is such that no relay strategy beyond basic forwarding makes sense for energy-constrained nodes.
Before directly addressing this issue, we first observe that the benefits of cooperation go beyond reducing transmit power. For example, COPE increases throughput by using channel bandwidth more efficiently. In high load scenarios, these benefits should not be ignored, even if energy consumption for listening to overheard packets increases.
Returning to the issue at hand, the benefit of reduced transmit powers must be balanced against the receive energy costs at the relays. To examine this issue, we compare the data-handling capabilities of some familiar devices and the energy costs that are incurred. These comparisons are based on current practices and technology for transceiver design, as opposed to any fundamental limits on the energy cost of computation. Nevertheless, the following discussion highlights that receive energy is an issue meriting further study.
We begin with the ubiquitous mobile phone that advertises a talk time on the order of 4 h. As the optimized sleep mode of the phone enables a battery lifetime of several days in the absence of talk, we can conclude that the energy used for non-communication tasks is negligible. Thus, in the simplest analysis, a phone can consume its battery in cooperation if it spends 4 h relaying the packets of other nodes.
Continuing, consider a Motorola RAZR with a 3.7 V battery rated at 740 mAh that stores 9.8 kJ. During a call, the phone is either transmitting and/or receiving data bits at roughly 104 b/s. In fact, the phone may be transmitting and receiving simultaneously but we ignore this factor of two since it is roughly cancelled by a 40% voice activity factor [63]. Thus, in 4h (14,400s), 1.44 x 108 bits (18 MB) are communicated. In using 9.8 kJ, the energy cost is 544 J/MB.
Now consider the 802.11 WLAN interface. An Atheros whitepa-per [10] found that typical WLAN interfaces consumed 2-8 W while actively communicating. As the actual transmit power of a 802.11 device is in the range of 20-100 mW, signal transmission uses only a small fraction of the power budget of even a 2 W WLAN interface. That is, the power required for signal processing dominates the transmit power and thus the power consumption is roughly the same for transmission and reception. In fact, this should not be surprising as the energy consumption of receiver processing is closely tied to the bandwidth expansion associated with the receiver’s soft-symbol interfaces as discussed in Section. 6.4. In any event, it follows that a dedicated WLAN interface communicating at 3 MB/s consuming 3 W power will operate at 1 J/MB.
Mobile phones and 802.11 laptops represent two extreme points in mobile communication. The low energy/bit efficiency of the mobile phone derives from communication distances on the order of 1 km, roughly 10-100 times typical WLAN communication distances. With an a = 3 path-loss exponent, the resulting transmit energy per bit for the mobile phone is higher by a factor ranging from 103 to 106. In the case of a cellular phone, a mitigating factor is that a cellular base station can use a large directional antenna to reduce the relative loss by perhaps a factor of 10 or more. Thus, our initial estimate that a cellphone may require 500 times the Joules/bit of a WLAN interface is in the right ballpark.
The cellphone’s dramatically higher per-bit energy costs arise because cellphones and WLAN devices operate in different regimes.
The required transmit power dominates a cellphone’s energy budget, while signal processing dominates for WLAN devices. These power consumption regimes are well recognized. In the context of Berkeley motes, [72] models the energy cost per bit for a reliable 1 Mb/s link over a distance d with path-loss exponent a = 2 by a transmitter cost of
where
is the energy dissipated in the transmitter electronics and
scales the required transmit energy per bit. In addition, as signal detection is more complex than signal synthesis, the receiver was found to expend Erx = 1.08 J/MB in signal processing.
We observe that bit processing costs are comparable for the mote and the 802.11 WLAN. This is not surprising since processing costs are dictated by current circuit technology. Thus, with respect to energy consumption per bit, wireless links can be thought of as occurring in two distinct flavors:
• Long Distance Links: The transmit power requirements dominate.
• Local Links: Signal processing dominates and the power consumption is proportional to the communication rate (sum of bit rate inflow and outflow).
For the mote model in [72], transmission and processing energy costs are equal at the transition distance of
For other systems and environments, the transition distance depends on system factors such as the receiver design and antenna size and environmental factors such as the path-loss exponent. Nevertheless, the transition distance is typically on the order of a hundred meters for common systems. For example, [36] studied the energy efficiency of MIMO and cooperative MIMO systems. For each signaling strategy, there is a long distance regime, typically starting around 50 m, in which the energy cost of additional receiver complexity is more than compensated by the reduction in transmitter power.
Advances in circuit technology are likely to reduce the transmitter and receiver bit processing costs and Erx. Nevertheless, per-bit transmit energy costs will increase as da and thus there is always a long-distance regime in which transmit power is the dominant cost. These energy consumption figures demonstrate that irrespective of the technology and communications system (cellular, WLAN, motes, etc.), there is always a long-distance regime in which conserving transmit power is the primary consideration. In these scenarios, one cannot ignore the savings in transmit power afforded by cooperative relaying.
