Advances in integrating autonomy with acoustic communications for intelligent networks of marine robots

Abstract

Autonomous marine vehicles are increasingly used in clusters for an array of oceanographic tasks. The effectiveness of this collaboration is often limited by communications: throughput, latency, and ease of reconfiguration. This thesis argues that improved communication on intelligent marine robotic agents can be gained from acting on knowledge gained by improved awareness of the physical acoustic link and higher network layers by the AUV's decision making software.This thesis presents a modular acoustic networking framework, realized through a C++ library called goby-acomms, to provide collaborating underwater vehicles with an efficient short-range single-hop network. goby-acomms is comprised of four components that provide: 1) losslessly compressed encoding of short messages; 2) a set of message queues that dynamically prioritize messages based both on overall importance and time sensitivity; 3) Time Division Multiple Access (TDMA) Medium Access Control (MAC) with automatic discovery; and 4) an abstract acoustic modem driver.Building on this networking framework, two approaches that use the vehicle's "intelligence" to improve communications are presented. The first is a "non-disruptive" approach which is a novel technique for using state observers in conjunction with an entropy source encoder to enable highly compressed telemetry of autonomous underwater vehicle (AUV) position vectors. This system was analyzed on experimental data and implemented on a fielded vehicle. Using an adaptive probability distribution in combination with either of two state observer models, greater than 90% compression, relative to a 32-bit integer baseline, was achieved.The second approach is "disruptive," as it changes the vehicle's course to effect an improvement in the communications channel. A hybrid data-and model-based autonomous environmental adaptation framework is presented which allows autonomous underwater vehicles (AUVs) with acoustic sensors to follow a path which optimizes their ability to maintain connectivity with an acoustic contact for optimal sensing or communication.

An Acoustic Network Navigation System

“…In acoustic‐based networks, the desired throughput is almost always above the available capability. To optimize the channel usage, the CMRE acoustic network uses a priority‐based queuing system (Schneider, ). The message priority is a combination of two parameters: the importance of the message and its time to live.…”

Distributed autonomy and formation control of a drifting swarm of autonomous underwater vehicles

“…In The underwater domain places unique constraints on swarm robotics and swarm intelligence, the primary limitation being that of communication. Underwater acoustic communication is generally limited by low bandwidth and intermittency due to the nature of the medium, in which multipath propagation, high ambient noise, and strong signal attenuation results in inter-symbol interference and low data rates [58]. A swarm system introduces an additional issue of message collisions, due to the number of agents needing to communicate.…”

“…chip-scale atomic clocks) and acoustic pingers; bearings between neighbours can be estimated via an acoustic hydrophone array (e.g. a tetrahedral 3D array) and sufficient signal processing, or alternatively by using vector sensors; and unique IDs can be communicated between two vehicles either by using an acoustic modem and compressed encoding/decoding schemes [58], or via narrowband acoustic pingers with a unique frequency for each vehicle. In general, when investigating swarm formation control for the underwater domain, it is obvious that control strategies that minimize the amount of information passed between agents are highly advantageous.…”

Visual summaries for low-bandwidth semantic mapping with autonomous underwater vehicles

“…The classification masks exhibit high redundancy and covariance so they can be compressed at high rates. These data can be transmitted using very little bandwidth with the techniques presented in [25] and [24].…”