AUV Swarms for monitoring rapidly evolving ocean phenomena

Abstract

Autonomous Underwater Vehicles (AUVs) have been commonplace in ocean science for years, taking the place of labour intensive and expensive ship operations. However, certain ocean phenomena, particularly those that change rapidly and require persistent monitoring over a large area, present challenges for AUVs, typically slow and energy restricted vehicles. Phenomena such as Harmful Algal Blooms (HABs) and oil spills fall into this category and have the potential to cause millions of dollars worth of damage to marine ecosystems, fisheries, human health and even tourism in coastal regions.

The emergence of a new class of micro AUVs allows much larger numbers of vehicles to be deployed than previously possible, making an AUV swarm (a large group of simple, small and low-cost robots) a feasible possibility. Due to the large number of vehicles, a swarm can resolve the temporal and spatial scales required to produce a cohesive data set of rapidly changing ocean phenomena. However, implementing an AUV swarm gives rise to the typical challenges associated with AUV operation as well as a new challenge, networking vehicles in a scalable manner.

Communication between AUVs is a significant challenge, satellite and WiFi methods only work on the surface and while acoustic communication provides a solution to underwater communication, it suffers from low bandwidth, high latency and high packet loss. This work focusses on subjecting established swarming algorithms to the communication constraints typical of AUV operation, evaluating whether the swarm is able to achieve an acceptable level of coordination.

In order to evaluate the performance of the swarm, an individual AUV model is developed based on the ecoSUB, a micro AUV suitable for swarm deployments. This model is then validated against vehicle deployments, highlighting factors unaccounted for in the model. In order to simulate a swarm, satellite and acoustic communication are added to the simulation, allowing vehicles to communicate and communication constraints to be imposed.

Firstly, an emergent flocking algorithm is evaluated, aiming to unify the vehicles as one group travelling in the same direction at the same speed. Swarms of 5 to 20 vehicles are tasked with following the flocking algorithm when subject to varying acoustic communication constraints. The results show a good level of flocking coordination across most swarm sizes and communication constraints, with the swarm consistently performing better than a single vehicle when completing a hill climb of a sensory function.

Secondly, a coverage control algorithm is evaluated, aiming to control the distribution of vehicles across an area, important when data is required at specific spatial scales in order to be most informative. Vehicle-in-loop experiments introduce a real vehicle to the simulated swarm and show the disturbances of vehicle drift and sensor noise to have little effect on the performance of the swarm. In simulation, the scalability of the algorithm is evaluated for increasing swarm sizes and communication constraints. Operating on satellite communication alone, the swarm is unable to achieve a stable distribution. Performance improves considerably with the introduction of acoustic communication, resulting in a stable distribution and a well coordinated swarm.

This work concludes that emergent flocking and coverage control are both suitable distributed control mechanisms for an AUV swarm, given acoustic communication is possible. Performance decrease with increasing swarm size and communication constraints is observed in performance metrics across both behaviours. However the performance decrease is not significant enough within the practically feasible ranges considered to state a limit of scalability.

More information

Identifiers

Catalogue record

Export record