11.2 Future Research Directions

The frameworks proposed in Chapters 4 and 5, which compute the throughput bound and capacity of OR need to find all the feasible concurrent transmitter sets, which is a NP-complete problem. How to find, efficiently, a good subset of all the CTSs to approach the optimal solution within a controllable gap could be an interesting topic. Some heuristic algorithms similar to that in Tang et al. (2007), or a column generation technique (Zhang et al. 2005) may be adopted to serve this purpose.

In multiradio multichannel networks, it is interesting to investigate distributed algorithms to solve the channel assignment and candidate selection issues. As we discussed in Chapter 5, there is a tradeoff between spatial diversity and multiplexing when applying OR in multiradio multichannel networks. Designing an efficient algorithm that leads to near-optimal channel assignment and candidate selection is desirable.

Routing metrics with various performance objectives, such as maximizing throughput, minimizing delay, and maximizing energy efficiency, can be studied and the tradeoff between conflicting goals can be analyzed and considered for OR.

Another potential direction for research is to investigate further the error of link quality (PRR) estimation and its impact on the OR performance in different types of networks, and design protocols accordingly that are robust to estimation error. We plan to break down this task into three subtasks.

First, geographical routing has been studied extensively in the literature in networks where location information is available to the nodes, which is true in many applications of multihop wireless networks. Geographic opportunistic routing has been proposed as an efficient routing scheme in such networks. In GOR, the Euclidean distance between nodes is known and can be used as the cost function in routing. We can start with GOR in wireless sensor networks where the distance is a fixed value and not affected by the link estimation error. We can design cost functions that are less affected by PRRs but represent the spatial diversity along the path.

Second, a local OR decision depends on OETT and EMT in Chapter 4. Opportunistic ETT only depends on the local PRRs whereas EMT depends on remote PRRs through Dis. Thus the impact of link estimation error is propagated through the network by Dis. For very dynamic networks, such as mobile ad hoc networks and vehicular networks, the link condition may change very fast, which may diminish the benefit from our optimization based on link error estimation. In our second step of understanding the impact of estimation error on routing performance, we will study Dis that are less sensitive to such changes. One option is to use the cost based on the traditional routing, which is less affected by the link estimation error than OR. The goal is to mitigate such impact from remote nodes and to focus on the impact on local estimations. Another option is to develop on-demand protocols (Johnson et al. 2001; Perkins and Royer 2001). Like multipath on-demand routing (Marina and Das 2001; Ye et al. 2003; Zeng et al. 2005), multiple replies can be enabled from the destination and nodes learn its local spacial diversity opportunity and report it in the reply messages. Spacial diversity along the paths can then be taken into consideration in routing decisions.

Third, after understanding the impact on each local decision, we will then extend the investigation to the whole paths. This study will be in a relatively stable setting such as sensor networks and mesh networks. We may adopt ideas similar to fisheye state routing (FSR) (Pei et al. 2000), which allows multilevel routing information exchange depending on the distance to the destination. The focus is to control the routing overhead while trying to take advantage of OR and path diversity on a larger scale. This study will help us to gain a deeper understanding of the OR and the capability of gaining performance benefits in the face of inaccurate link-quality estimation. We believe that the theoretical results and insights from this research will be valuable to the research community and crucial to the design of practical and efficient OR protocols that approach optimal performance.

Other than performance, security is another major concern in multihop wireless networks. Opportunistic routing, due to its indeterministic nature, should be more robust to many attacks aiming to disrupt routing and data forwarding functions. Generally, OR is more resilient than single-path traditional routing under packet-dropping attacks. However, in some cases, single-path routing may be better—for example, where the path length is short. It is worthwhile studying in depth the theoretical foundations of opportunistic routing under different networking configurations and attacking models and quantifying the damage or performance degradation that would result. It will be valuable to design secure OR protocols and integrate them into existing security framework to provide more robust and more secure information delivery service.

With recent advanced wireless communication technology, such as MIMO (multiple-input multiple-output), OFDM (orthogonal frequency-division multiplexing) and cognitive radio, it will be interesting to study how opportunistic routing can be integrated with these physical layer technologies to optimize network performance.