【正文】 RELATED WORKThere are few papers on measurementbased assessment of cooperative relaying schemes. Bradford et al. [6] measure and evaluate decodeandforward relaying schemes in the lab. Measurements are performed with fixed distances between nodes with very low mobility and RF shields influencing radio channels. Kyritsi et al. [7] measure the performance of cooperative relaying in an indoor office scenario with two access points and two devices moving on predefined paths. Valentin et al. [8] propose and implement a medium access control (MAC) protocol for mobile cooperative are carried out in a railroad scenario, where devices move on an oval shaped railroad with low mobility. Gonzales et al. [9] pare the achievable data ratios of relaying schemes in a scenario, where indoor mobile stations municate to an outdoor base station. In summary, none of these papers analyzes cooperative relaying in a realistic outdoor scenario with high speed vehicles.III. EVALUATION METHODOLOGYMeasurements are performed in a cartocar munications scenario, where three cars are equipped with WARPboards serving as sender S, destination D,andrelay R,respectively. The antennas are placed on the roofs of the cars. Each board is connected via Ethernet to a notebook to collect status packets and to track the positions of the cars with GPS sensors. Wireless munications is performed using the WARP orthogonal frequencydivision multiplexing(OFDM) reference design (version 12). A relaying protocoland a retransmission protocol are implemented by the authors at the MAC layer. Table I shows the transmission scenarios for the positions of S, D, R are studied:? Relay middle (RM): R is driving between S and D.? Relay last (RL): R is driving behind S and D.TABLE I: Transmission parametersParameterValueFrequencyBandwidth10MHzHeader length24bytesPayload length1024bytesModulationOFDM with QPSKAverage TX power of packet14/11dBm (full / half power)Peak transmission (TX) powe(full / half power)Fig. 1: Transmission cycle? Relay and destination in same car (RD): Two cars areused, one car acts as S, the other as both R and D, .,one car is equipped with two WARP boards。 the other two schemes employ halfpower packets (see Table I). As shown in Fig. 1, the MAC layer has been programmed to subsequently transmit? a fullpower packet Sf sent by S,? three halfpower packets Sh,Sh*, Sh** sent by S,? a halfpower packet Rh sent by R.The time period between two subsequent packet transmissions is δ =30ms.A packet is delivered, if it is received by the munication partner and the cyclic redundancy checks (CRCs) of both header and payload are valid. As shown in Fig. 2, a packet is delivered (a) by direct transmission if Sf is delivered to D,(b) with time diversity if at least one of the halfpower packets Sh orSh is delivered to D, and (c) using coopera tive relaying if Shis delivered to D,or Sh** is delivered to R and delivered to D. No packet bining is performance of time diversity and cooperative relaying depends on the time period between the ?rst and the second packet of the given scheme. This time is denoted by Δ in the following. To assess time diversity, we consider two packets Shand Sh* separated by Δ. For cooperative relaying, we consider Sh** and Rh separated by Δ. Packet delivery is evaluated as a function of Δ ranging from 30ms to 30 s, which means that Δ∈T , T := {(5i +1)δ| i =0, 1,..., 199}.Note that the time diversity scheme also bene?ts from spatial diversity due to movements of the cars. In each environment about 100 000 packets are assess the statistical signi?cance of the measurements,we uniformly and randomly select 2 000 transmission cycles from the data set into a subset B and evaluate this set. This procedure is repeated ten times. We show the mean of these ten values and the 10% and 90%quantiles.IV. TEMPORAL CORRELATION OF PACKET RECEPTIONThe temporal correlation of packet reception of a given link is a key factor for the performance of both time diversity and cooperative relaying (which is spacetime diversity). The idea behind time diversity is that a retransmission may be successful if a ?rst transmission failed due to changes of he channel. Informally speaking, a large positive correlation means that little changes of the channel can be expected, which in turn makes it unlikely that the second transmission succeeds if the ?rst one failed. High correlation will lower the performance of time diversity. Therefore, temporal correlation can be used as an indicator for the probability of successful transmission in time diversity and enables us to choose a suitable interval Δ between the diversity packets.We analyze the correlation of halfpower packets for intervals Δ∈T .Let T := {ti | i =1,...,κ} denote the set of κ time instants of B when S sends halfpower packets. We introduce the binary variable Rti=1 if halfpower packet sent at ti is delivered to R0 else We evaluate the sample correlation ρ( D, R,Δ) with D :=(Dt1 ,Dt2 ,...,Dtκ) and R :=(Rt1 ,Rt2 ,...,Rtκ ) as ρ( D, R,Δ)=1(T1)sDsRt=TDtD(Rt+ΔR),where R and D are the arithmetic means, sR and sD thesample variances, and T the subset of T contain ing all time instants when S sends halfpower packets Sh at times t ∈ T for which t ≤ tκ ? Δ holds.Fig. 3 shows correlations for both suburban and highway environments. The auto correlation ρ( D, D,Δ) shown in (a) is highly positive if the time span Δ between packets is short。for Δ ≥ s,the correlations of highway and suburban environments are similar. For both environments, such positive correlation will reduce the performance of time diversity.Fig. 3: Temporal correlation. Results are for the RD case, but the correlations are similar for the RM and RL cases.Fig. 3 (b) shows the correlati