This paper presents a comprehensive study on random access (RA) design for multi-channel Ambient IoT (A-IoT) networks enabled by uplink frequency division multiple access (FDMA). We introduce and unify the three-step RA procedure (RAP) specified in ongoing 3GPP standardization, which serves as a useful reference for future research. To fully exploit the multi-channel capability, we propose enhanced frame structures, including a novel multi-round Msg1 collection (MRMC) scheme, in which Msg1 carries a randomly generated ID from each device and is transmitted in the first step of the three-step RAP. We also develop new analytical models that capture the unique characteristics of multi-channel A-IoT, providing closed-form expressions for RA efficiency that serve as a foundation for network optimization. Furthermore, we design online control algorithms that estimate the device population in real time and dynamically adjust frame configurations for optimal operation under device uncertainty. The proposed Boosted Estimation with Early Frame Termination (BE2FT) algorithm effectively handles bursty traffic conditions and improves resource utilization. Extensive simulations validate the accuracy of our analytical models and demonstrate significant gains in RA efficiency and service time. The proposed scheme achieves performance close to the ideal performance, where the number of devices is assumed perfectly known and the frame length is optimally configured.

In machine type communication (MTC), diverse applications requiring high reliability and low latency lead to the consideration of ultra-reliable and low-latency MTC (uMTC) communication scenarios. Consequently, two-step random access procedure (RAP), as an alternative to the four-step RAP, is introduced into the 5th generation (5G) communication systems to reduce the unnecessary latency caused by the multi-round transmissions in the wireless medium. In this paper, we first analyze the delay performance of the two-step RAP based on which we further propose an algorithm to control the number of preambles allocated for the two-step RAP to meet a given average delay requirement. In particular, our proposed algorithm estimates the number of active devices in an online manner and controls the number of preambles to be allocated. Through extensive simulations, we show the effectiveness of our proposed algorithm in satisfying the delay requirement as well as minimizing the preamble resource usage.