This paper presents two new techniques employing bandwidth-efficient modulation to enable two partners to cooperatively convey their data to a common destination. Both approaches integrate relaying and channel coding, exploiting the fact that each partner knows the data relayed by the other partner to enhance the partner-to-partner link. The first approach is based on bit-interleaved coded modulation (BICM); a key innovation is the use of set-partition (SP) labelling that enables the receiving partner to demodulate and decode based on a sparse subset of the original constellation. The second approach is based on trellis coded modulation (TCM); the local and relayed data are multiplexed together prior to encoding, and the receiving partner prunes the trellis edges inconsistent with the known relayed data. In both approaches, the receiver at the common destination benefits in two ways - indirectly from the increased diversity made possible by the enhanced partner-to-partner link, and directly by exploiting the SP-labeling/trellis-pruning in its decoding protocol. Simulations indicate that the new techniques offer significant performance gain over conventional cooperation based on time-sharing - a gain of 2.50 dB for the 16-QAM BICM system and a gain of 3.0 dB for the 8-PSK TCM system, both at a frame error rate of 10-2.

This paper compares the performance of convolutional codes to that of LDPC block codes with identical decoding latencies. The decoding algorithms considered are the Viterbi algorithm and stack sequential decoding for convolutional codes and iterative message passing for LDPC codes. It is shown that, at very low latencies, convolutional codes with Viterbi decoding offer the best performance, whereas for high latencies LDPC codes dominate - and sequential decoding of convolutional codes offers the best performance over a range of intermediate latency values. The "crossover latencies" - i.e., the latency values at which the best code/decoding selection changes - are identified for a variety of code rates (1/2, 2/3, 3/4, and 5/6) and target bit/frame error rates. For sequential decoding, both blockwise and continuous resynchronization procedures are used to allow the decoder to recover the correct path. The results indicate that sequential decoding substantially extends (beyond what is possible with Viterbi decoding) the range of latency values over which convolutional codes prove advantageous compared to LDPC block codes.

A recent paper by the authors demonstrated how two users (or partners) could cooperate to convey their data to a common destination through trellis pruning. With this technique, both “local” and relayed data are multiplexed together prior to modulation; then, the partner's decoder exploits its knowledge of the relayed data by pruning the code trellis to effect a lower rate (more robust) partner-to-partner link. This paper extends this technique to larger clusters of users cooperating to convey their data to a common destination. The protocol is first demonstrated for n=3 users, and a significant performance gain is demonstrated over time-sharing in a Rayleigh fading environment. Then, the general protocol is derived and conclusions regarding performance are presented.

This paper introduces a new approach to cooperative communications. Consider two partner nodes that cooperate to convey their data to a common destination, each transmitting not only its own "local" data but also acting as a relay for its partner, thereby effecting spatial diversity. In the proposed scheme, each partner multiplexes local data with relayed data and convolutionally encodes the resulting bitstream prior to transmission; the other partner then uses its knowledge of the relayed data to prune edges from the code trellis, thereby decoding at a lower (and more robust) rate. The destination observes two copies of each source's data - one transmitted as local information, the other transmitted as relayed information - and employs iterative decoding to recover the same. The resulting performance is superior to that of cooperation schemes based on time sharing - for example, a gain of 4.1 dB is observed at a frame-error rate of 10-3with memory M = 3 codes - as well as schemes employing the newer algebraic superposition approach of Xiao et al., with trellis pruning enjoying an advantage of 0.55 dB to 1.04 dB over algebraic superposition, depending on M. The new approach requires some additional decoding complexity at each partner node, but the decoding complexity at the destination node is kept constant.

This paper introduces a new approach to cooperative communications. Consider a system wherein two partner nodes cooperate to convey their data to a common destination, each not only transmitting its own “local” data but also acting as a relay for its partner, thereby effecting spatial diversity. In the proposed scheme, each partner interleaves local data with relayed data and convolutionally encodes the resulting interleaved bitstream prior to transmission; the other partner then uses its knowledge of the relayed data to prune edges from the code trellis, thereby decoding at a lower (more robust) rate. The destination observes two versions of each source's data - one transmitted as local information, the other as relayed information - and employs iterative decoding, in effect decoding the data as if it were encoded via a “distributed” turbo code. The resulting performance is superior to those of cooperation schemes based on time sharing and signal space superposition, and it generalizes the algebraic superposition scheme of Xiao et al.

This paper compares direct and relayed transmission in a simple four node configuration with a source, a destination and two parallel relays. The channel model assumes additive white Gaussian noise, attenuation from path loss, and slow Rayleigh fading. The two parallel relays use the Alamouti code to relay data to the destination. Using outage probability as a metric, we investigate the relative advantage of relaying as a function of relay placement and the required end-to-end spectral efficiency. We first quantify the benefit derived from adding a second relay to a one-relay network; we then calculate the "critical" rate, above which direct transmission always yields a lower outage probability. It is seen that the critical rate for a two- relay system (with diversity combining) is only modestly higher than that of a one-relay system - suggesting that the addition of more parallel relays is not an effective means of increasing the range of end-to-end rates over which relaying provides an advantage over direct transmission.

This paper compares routing strategies in a simple four node network consisting of a source, a destination, and two parallel relays. Relay-to-relay communication is not considered, so there are three possible routing strategies: (1.) direct (source-to-destination) transmission; (2.) two-hop transmission that makes use of exactly one relay (i.e., the better-placed relay); and (3.) two-hop transmission wherein both relays are used during the second hop. The channel model assumes additive white Gaussian noise and attenuation due to path loss; for two-hop routing, time division multiple access is assumed for interference mitigation. Using simple capacity arguments, the paper first compares two-hop routing through one relay and two-hop routing through two relays and characterizes the relay placements for which one mode is preferable to the other. Then, two-hop transmission exploiting both relays is compared with direct transmission. It is shown that if the desired end-to-end rate is above a critical spectral efficiency, it is always preferable to transmit the signal directly no matter where the two relays are located.

This paper compares one-hop and two-hop routing in a simple three node network consisting of two mobiles and a base station. The channel model assumes additive white Gaussian noise, attenuation due to path loss, and slow Rayleigh fading; the network is assumed to use time division multiple access to ameliorate interference for two-hop routing. The outage probabilities of the two schemes are compared to derive conclusions about the relative advantages of multi-hop (two-hop) routing as a function of required end-to-end spectral efficiency. This analysis is carried out under two different assumptions regarding the allocation of power between the mobiles - one scenario in which a total sum power is allocated optimally and another scenario in which each mobile is allocated the same fixed power. The analysis also includes two different assumptions about the receiver's operation in two-hop mode - one in which the receiver bases its estimate of the data solely on the signal transmitted by the relay, and another in which the receiver uses code combining to exploit both the relay-transmitted signal and the source-transmitted signal. For the case in which the receiver ignores the source-transmitted signal, it is shown that one-hop routing outperforms two-hop routing when the required end-to-end spectral efficiency is above R* = log2(2alpha-1-1), where alpha is the path loss exponent. When code combining is employed at the receiver, this "critical rate" is substantially increased.