E-MAC: An evolutionary solution for collision avoidance in wireless ad hoc networks

Abstract

Transmission collision is a main cause of throughput degradation and non-deterministic latency in wireless networks. Existing collision-avoidance mechanisms for distributed wireless networks are mostly based on the random backoff strategy, which cannot guarantee collision-free accesses. In this paper, we design a simple collision-avoidance MAC (E-MAC) for distributed wireless networks that can iteratively achieve collision-free access. In E-MAC, each transmitter will adjust its next transmission time according to which part of its packets suffering from the collision. And the iteration of this adjustment will quickly lead group of nodes converging to a collision-free network. E-MAC does not require any central coordination or global time synchronization. It is scalable to new entrants to the network and variable packet lengths. And it is also robust to system errors, such as inaccurate timing.

Introduction

In distributed wireless networks, the overlap of more than two interfering nodes׳ transmission will result in transmission collisions. And it is a common phenomenon due to the broadcast nature of wireless communication and the contention for shared wireless medium. Transmission collision can be classified into two categories: the synchronized collision and the hidden node collision. The synchronized collision occurs when no less than two nodes happen to start their transmissions simultaneously, as illustrated in Case A of Fig. 1. And this usually happens in single-hop wireless networks, where more than two nodes happen to select the same backoff counter to transmit after detecting that the wireless medium is idle. Unlike the synchronized collision, in hidden-node collision, the colliding packets can be overlapped at any stages, including all the cases from A to D in Fig. 1. And the hidden-node collision is the dominating collision scenario in multi-hop networks, where there exist hidden nodes. Zhao et al., 2013, Zhao et al., 2010) modeled hidden-node collision and analyzed its effect on end-to-end throughput demonstrating that hidden-node collision can decrease the throughput by more than 20% in high density networks.

When collision occurs, the packets cannot be correctly received and have to be re-transmitted, resulting in throughput degradation, non-deterministic latency and wastage of radio resources such as channel bandwidth and energy. To reduce the probability of collisions, currently popular collision avoidance mechanisms for wireless networks are mostly based on the random backoff strategy (such as the Binary Exponential Backoff in 802.11 DCF and its improvements; Wang and Zhuang, 2006, 2008; Van Nee, 2011; Tuysuz and Mantar, 2014; Wang et al., 2004; Madhavi and Rao, 2015). However, these mechanisms have two drawbacks. First, it cannot guarantee collision-free accesses, and second, it is not efficient due to the nature of random backoff. The situation is even worse in high-speed wireless networks such as IEEE 802.11 ac standards where the PHY rate reaches as high as 1 Gbps (Van Nee, 2011).

In this paper, we propose the E-MAC that uses elaborate, instead of random, backoff mechanism to iteratively achieve collision-free access. The basic idea of E-MAC lies in that, every node transmits at most once in a given time cycle, and if one node experiences a collision, it will adjust its transmission time in the next cycle according to which part of its packets suffering from the collision. Specifically, if the front part of the packet is collided, it will transmit latter in the next cycle; on the other hand, if the back part is collided, it will transmit earlier in the next cycle. We modified the 802.11 MAC to implement this idea and proved the effectiveness of the modification via both theoretical analysis and computer simulation. The main characteristics of E-MAC are as follows:

• Running E-MAC, the network can quickly converge to collision-free access. In the initialization process, the network efficiency is no worse than the traditional random backoff schemes, and after achieving the collision-free access, the network efficiency can approach 100%.

• E-MAC is a totally distributed protocol, which requires neither a central coordination nor global time synchronization. Furthermore, it does not rely on the prior knowledge of channel states (for instance, the idle time slots and the busy time slots in the past transmission period).

• E-MAC is a dynamic adaptive protocol, which is resistant to possible system errors (e.g., inaccurate time slot split and time slot drift) and scalable to variable packet lengths and large number of network members.

The rest of this paper is organized as follows: Section 2 reviews related literature on collision avoidance protocol. Section 3 introduces the system model and defines the problem of this paper. Section 4 presents the proposed E-MAC protocol, followed by the theoretical analysis on its performance and then Section 5 explains the implementation issues. The performance of the proposed protocol is evaluated in Section 6 with computer simulations. Finally, Section 7 concludes this paper.