In this paper, a model is developed for a slotted time-division multiplexing system used for assigning transmissions over a set of time-varying channels. Two schemes, a "round-robin" and an adaptive multiplexing scheme, are studied for use with three standard automatic-repeat-request (ARQ) protocols, namely, stop-and-wait (SW), go-back-N (GBN), and ideal selective repeat (ISR). In round-robin multiplexing, transmission slots are assigned periodically to each channel, With the adaptive scheme, the multiplexer selects at each time slot the channel whose state is estimated to have the lowest retransmission probability. The throughputs of these ARQ protocols under either multiplexing scheme are analyzed and compared. The present paper contains a description of the system model and an analysis of the ARQ protocols under both multiplexing schemes. In Part II, a modification to GBN and selective repeat ARQ for reducing the detrimental effects of feedback errors is discussed and analyzed, followed by a throughput performance comparison of all ARQ and multiplexing schemes.

For pt.I see ibid., vol.46, no.1, p.41-51, 1998. A "postponed retransmission" (PR) modification to both go-back-N (GBN) and ideal selective repeat (ISR) automatic-repeat-request (ARQ) is discussed and analyzed. A scheme for reducing the number of states involved in the calculation for multiplexed GBN ARQ is also presented. Numerical results are then shown for stop-and-wait (SW), GBN, and selective repeat (SR) ARQ (as well as the PR modified versions), under both round-robin and adaptive multiplexing.

Automatic-repeat-request (ARQ) protocols have been analyzed for quite some time, but the issue of errors in the feedback channel has not received much attention. In some applications, such as digital mobile communications, this issue can be important. Accordingly, this paper examines the effect of feedback errors on the throughputs of the stop-and-wait (SW), go-back-N (GBN), and selective repeat (SR) ARQ protocols for the ease of a point-to-point channel under some feedback information assumptions. It is shown that the deleterious effects of feedback errors on the throughputs of continuous (e.g., GBN and SR) ARQ protocols can be greatly reduced by a simple modification in the retransmission operation, provided that the "complete state" of the receiver is sent back with each acknowledgment.

Six hard-decision combining schemes are considered. The probability of decoding failure P/sub F/ for the first two schemes can be easily obtained. For Scheme 3, an analytic expression for P/sub F/ is derived and for Schemes 4 to 6, a procedure is described which allows P/sub F/ to be evaluated numerically. A performance comparison of the six schemes is given.>