Underwater Acoustic Communications and Networking

Underwater communications are critical to cooperation between underwater assets and surface assets. As EM waves are not effective for medium-long range communication underwater, we use acoustics for underwater communications. The snapping shrimp noise dominated environment in Singapore waters limits the performance of many commercial communication systems. We have therefore developed our own acoustic communication system that is robust to non-Gaussian noise, and works well in both warm and cold shallow waters.

The ARL Modem

We have developed a communication technique for point-to-point medium range links in warm shallow waters. The technique employs OFDM to divide the available bandwidth into multiple sub-carriers. Each sub-carrier is narrow enough to lie within the coherence bandwidth of the channel. With the use of differential QPSK on each sub-carrier, we do not require explicit channel estimation or equalization. By using block interleaved FEC coding across all sub-carriers and multiple symbols, we harness gains from the time and frequency diversity available. The frequency diversity gives us robustness against frequency-selective fading, whereas the time diversity gives us robustness against transient events such as snaps made by snapping shrimp. The technique also includes the use of blocks such as sign correlator [16] and 1-norm Viterbi decoder [9] to ensure that the signal processing is near optimal in impulsive snapping shrimp noise.

An initial version of the technique was tested with off-line processing in Singapore waters [18], Woodshole harbor (unpublished, 2005) and Italian waters (unpublished, 2009). In 2006, a real-time implementation of the algorithm was developed on a combination of a FPGA, DSP and a single-board Linux host. This implementation has been extensively tested in Singapore waters in a frequency band of 50-75 kHz at ranges up to 1.5 km to yield link rates of up to 15 kbps. A 25-37 kHz version has also been implemented and tested to enable longer range communication.

A new version of the ARL modem is currently being developed. The hardware used for this modem has more computational capacity and therefore is able to support more advanced signal processing such as impulse noise cancellation, partial sparse equalization and Doppler compensation, as well as stronger FEC codes. The modem will also support multiple modulation schemes concurrently to allow links with different performance characteristics (such as communication range, packet loss probability, link data rate, etc) to be operated simultaneously. This support may also provide a path to interoperability with other commercial and research modems.

The ARL modems are compliant with the UNA architecture, and support the physical layer FAPI interfaces.

25-37 kHz ARL modem in a test housing

Adaptive Link Tuning

The ARL modems support a large set of options such as number of sub-carriers, cyclic-prefix length, cyclic-suffix length, number and positions of null carriers, peak-to-average power reduction, transmission power, detection preamble length, coding scheme (e.g. Golay, convolutional codes, LDPC, repetition codes), detection threshold, etc. When these options are correctly tuned for the channel, the performance of the modems is very good. However, with such a large number of options, the search for a good option set is not trivial.

We have developed data-driven techniques for adaptive real-time tuning of the options. One of the basic techniques has been demonstrated successfully during field trials. The technique uses a set of test packets that are transmitted on an idle link to help find a good option set to be used once data is available. More advanced versions of the technique are currently being tested in simulation and will be tested in the field soon. An documented interface allows new algorithms for link tuning to be implemented easily on the modem.

Link tuner framework

Media Access Control

Once we have a good communication link, the next natural step is to allow more than two modems to be able to share the medium. We explore medium access control (MAC) protocols and related aspects for an underwater acoustic network (UAN) primarily for a distributed topology. Though there are differences such as long propagation delays in UANs as compared to terrestrial networks, the basis for network protocol design is very similar in both types of networks. Traditional forms of multiple access protocols include CDMA, TDMA and FDMA. As shown in [5]; in a distributed topology, CDMA and FDMA require full-duplex and multi-channel functionality to have similar performance as TDMA. Dynamic forms of TDMA are well suited for distributed adhoc networks. Medium access with collision avoidance (MACA) based protocols are shown to be closely related to the dynamic TDMA protocol in [5] , and is the main area for MAC research in ARL.

Non adhoc TDMA was one of the first MAC protocols implemented for ARL modems including those used in the STARFISH AUVs. For few number of nodes in a single collision domain, TDMA provides a simple and robust solution to the MAC problem. It however requires precise time synchronization between nodes. The implementation and testing of this protocol is primarily an engineering effort.

Most of the focus in adhoc protocols is on MACA based protocols. Such protocols were investigated by ARL as early as 2005 [19] and found to be very promising for adhoc underwater MAC. There are a number of open research problems in this area including analytical performance characterization, protocol performance enhancement techniques etc. A number of publications such as [11], [5], [1] and [4] capture our progress in exploring this protocol family.

MACA Protocol Outline

Standardization Initiatives

ARL also contributes to standardization efforts such as NURC's JANUS initiative. ARL, in collaboration with MIT and WHOI, proposed an inter-layer communication standard called UNA in 2006 [15]. We recently also proposed how MAC could be implemented in a network with hetergeneous assets [5].

In [4], ARL highlights the importance of using an unified simulation and implementation framework for MAC protocol investigations. The use of such a common framework ensures accurate simulation and a quick time-to-field-testing.

Unified framework for simulation and implementation

Trials and Experiments

ARL conducts a host of field trials to validate simulation results and develop a strong understanding of the shallow water communication channel. This has led to validated shallow water channel models [10] that may be used for accurate simulation. It has also allowed us to test protocols and schemes at various layers in marinas and open ocean environments.

Most preliminary testing is done at the RSYC marina:

MAC protocol experiments are heavy in preparation, but relaxed in execution!

Multiple modems are deployed across the marina and linked on the surface via wifi for easy monitoring of the tests

Typical sea trials involve multiple boats to form a network:

A network experiment at sea

Current Theoretical Research Areas

Although we spend a significant amount of effort on experimental research and developing systems that can be tested at sea, we also engage in theoretical research that may feed into underwater communication systems years down the road. Some of the current areas of theoretical research include:

• Coding for channels with multi-dimensional symmetric alpha-stable noise
• Performance limits of networks with propagation delays
• Rateless and network coding for underwater networks
• Exploration v/s exploitation trade-offs for optimal link tuning

Related Publications