Overview

In this section, we provide the required background on both, channel hopping and op-portunistic routing in low-power WSNs. Next, we introduce the basic concepts of MOR.

2.3.1. Channel Hopping Strategies in WSNs

Regarding the selection of channels, channel hopping strategies fall into two categories:

“whitelisting” and “blind hopping” [WMP09]. In whitelisting, neighboring nodes agree on which channels to use at what point in time for their communication. In blind channel hopping, nodes do not know which channels their neighboring nodes use at what point in time. To establish communication, nodes uniformly hop over all utilized channels, i.e., up to16 radio channels in IEEE 802.15.4[ISA11].

Practically, there are two types of channel allocations in multichannel communication, i.e., static channel allocation and dynamic channel allocation [HXS⁺13]. Depending on the scenarios, dynamic channel allocation can be more effective if the interference condi-tion is changing dynamically over time. It, however, often performs complex rendezvous algorithms, thus, resulting in non-trivial communication computing overhead. To bal-ance the performbal-ance and the computing overhead of the sensor node, MOR chooses to use static channel allocation.

The main goal of any channel hopping scheme is to increase robustness towards interference. We observe three approaches of channel hopping strategies in wireless communication: fast channel hopping, slow channel hopping, and hybrid channel hop-ping [HXS⁺13]. Fast channel hopping switches to a new channel in each time slot. Fast channel hopping is used in a number of applications and standards in order to improve secrecy and to make the system more robust against jamming or interference. For ex-ample, Bluetooth and WirelessHART [Fun06] employ fast channel hopping. Meanwhile, this approach increases the overhead for a packet transmission, i.e., frequent channel switching makes a device consume energy faster than others. Slow channel hopping stays for multiple continuous time slots on a single channel before switching. Compared to fast channel hopping, slow channel hopping generates less latency when two devices need to rendezvous on a common channel. Hybrid channel hopping combines both fast and slow channel hopping, where fast channel hopping improves the robustness towards interference and slow hopping allows for fast rendezvous.

Generally, MOR exploits hybrid channel hopping scheme. Duty-cycled sensor nodes perform fast channel hopping to ensure robustness towards the interference. That is, they switch to a new channel in a short time slot so as to avoid keeping using a interfered channel for rather long time. An always-on node (i.e., the sink), which does not go to sleep mode at all, employs the slow channel hopping scheme to guarantee the fast rendezvous of the last-hop neighbors. In this case, whenever there comes a packet from last-hop neighbors to the sink, the sink can capture and receive it in at least one

“good” channel, simply because that the last-hop neighbors hop to a new channel more frequently than the sink does.

2.3.2. Opportunistic Routing in WSNs

Approaches to opportunistic routing in duty-cycled WSNs differ from traditional unicast, where packets are addressed to one specific neighbor. In traditional unicast, as shown in Figure 2.3(a), if node 1 has a data frame to send, then it keeps sending a data frame via a reliable link. A receiver, on the other hand, wakes up and detects the data by a Clear Channel Assessment (CCA). In IEEE802.15.4, the MAC layer employs the Carrier Sense Multiple Access with Collision Avoidance (CSMA/CA) mechanism. CCA is used in the physical layer to determine the channel occupancy [ISA11]. Generally, a CCA performs Energy Detection (ED), or Carrier Sense (CS), or a combination of both. CCA aims to report a busy channel upon detecting any energy above a preset ED threshold.

Afterwards, the receiver sends an acknowledgment back to the sender, i.e. node1. Node 2 then sends the data frame to the destination, i.e., node 3. Node 3 wakes up, detects the data by a CCA, receives the data frame, and sends back an acknowledgment. In this case, the routing set is built based on the link quality. That means one node selects its

next-hop forwarder from the neighboring nodes based on the link quality.

Opportunistic Routing for Wireless sensor networks (ORW) [LGDJ12] is an oppor-tunistic routing scheme for duty-cycled WSNs. ORW uses anycast addressing a one-to-any-one scenario where data packets are routed to any single member of a group of potential receivers. Consequently, data packets in ORW are forwarded by one of the neighboring nodes which (i) wakes up first, (ii) successfully receives the packet, and (iii) provides routing progress. As shown in Figure 2.3(b), in LPL-anycast, node 1 repeats sending the data frame regardless of the link quality. The next-hop node, who wakes up earlier, detects the data frame using a CCA, receives the data, and acknowledges the sender.

ORW is able to sufficiently reduce delay and energy consumption and improves the resilience to wireless link dynamics. Furthermore, Opportunistic RPL (ORPL) integrates the concepts of opportunistic routing with RPL [Win12], the standard protocol for low-power and lossy Internet Protocol version 6 (IPv6)-based networks. ORPL provides any-to-any and on-demand traffic. Both ORW and ORPL utilize the Expected Duty Cycles (EDCs) [LGDJ12] as the routing metric. When a node is selecting its next-hop forwarder from its neighboring nodes, EDCs of the neighboring nodes are used as a metric to compare. This allows the node to select the set of neighboring nodes in different hops that provide sufficient routing progress. Experimental results from testbeds show that ORW and ORPL outperform the state-of-the-art solutions including RPL and the Collection Tree Protocol (CTP) [GSC09] in terms of latency, power consumption, robustness, and scalability.

D D D

c

c

D D D D D

D D D D

c D

A A

A A 1

2 3

D: Data Frame A: Acknowledgement C: Clear Channel Assessment

c

(a) LPL-based unicast. One sender unicasts the data packet over a single channel to a neighbor based on a routing metric, e.g., link quality.

D D D

c

D D D

D A A

c c

c c

1 2 3

D: Data Frame A: Acknowledgement C: Clear Channel Assessment

(b) LPL-based anycast. One sender anycasts the data packet over a single channel to the neighbor who wakes up earliest thus, reducing the delay.

Figure 2.3.: Low-power listening-based unicast and anycast using the same topology as the one in Figure 2.2.

2.3.3. MOR in a Nutshell

MOR extends opportunistic routing with multichannel hopping to combine their key advantages: low latency and high energy efficiency of opportunistic routing with strong robustness to interference of multichannel hopping. Thus, MOR inherits the spatial and temporal diversities of opportunistic routing and additionally exploits the frequency diversity of multichannel routing.

MOR builds on ORPL: It employs the EDC routing metric [LGDJ12] and the integra-tion with RPL [Win12]. Addiintegra-tionally, unlike a number of synchronous MAC protocols for WSNs, e.g., [KSC08] and [IVHJH11], MOR is based on asynchronous Low-Power Listening (LPL). It, thus, does not lead to additional synchronization overhead within the network and efficiently operates its channel hopping without coordination overhead.

Moreover, MOR does not only transmit opportunistically, but also selects channels op-portunistically: For each listening and (re)transmission of the underlying MAC layer, MOR utilizes a new channel. For example, while in ORPL it takes multiple transmissions of the MAC on a single channel until one neighboring node wakes up and successfully receives the packet, MOR does each of these (re)transmissions on a different channel.

Overall, MOR extends the concept of opportunistic routing to the frequency domain.

That is, in MOR, the first node that wakes up on the rendezvous channel and successfully receives the packet, acts as a forwarder and, thus, provides the routing progress. We show in our experimental evaluations that MOR significantly improves robustness in the presence of interference when compared to other state-of-the-art protocols. In addition, we show that the duty cycle of MOR is only approximately0.3% higher when compared to our baseline protocol ORPL in interference-free scenarios.