All systems go

Autonomous underwater vehicles (AUVs) are essential to deep-water oil exploration, military reconnaissance and scientific expeditions, exploring hazardous marine environments where humans cannot travel. However, construction, maintenance and operational costs mean that if they fail to function during an exploratory voyage, thousands of pounds could be wasted.

Prof Sandor Veres from the School of Engineering Sciences at Southampton University is leading a European-funded project to increase the reliability of AUVs and reduce the risk of them malfunctioning or becoming lost.

'AUV reliability has several aspects,' said Veres. 'First, there is the hardware, which is constructed component by component. An oceanographic AUV may cost £1m as the components are expensive. Although tested individually, when assembled, it does not necessarily mean they will work together as planned. This project will use formal methods for hardware modelling.

'Second, we will look at the software, which controls the vehicle in the hazardous environments in which it operates.'



Operating time

The cost of missions using AUVs is an important aspect of the project, both in terms of hardware and operating time. The National Oceanography Centre, for example, has a six metre-long AUV that weighs over three tonnes and is worth more than £1m.

'When it's operating at sea, if there's a failure on board at a depth of 3km, it has to resurface at a speed of 2m/s, at an angle of 30º,' said Veres. 'It could take an hour to come up, then take a day to rectify the problem.

'The cost of a scientific mission is £1,000 an hour and, in the case of the oil industry, this cost can be higher, as well as the operation of the ship being more expensive. In those industries, it can be £2,500 an hour.

'So this project is also of interest to the insurance industry, which is undertaking the cover for these missions for various organisations, both in the oil industry and in science.'

The project aims to develop a methodology that could be applied to other autonomous vehicles operating in hazardous environments. Similar systems include unmanned aeronautical vehicles, satellites and bomb disposal ground vehicles.

The methodology will be developed in three stages: the team will develop a design optimisation tool for overall operational reliability, a fault assessment system and a human interface for daily operation.

'The problem with these systems are they are one-off and not mass-produced,' said Veres. 'There is no history of quality control and the people who are deploying them are not necessarily up to speed on how it works, so they make mistakes. We want to automate the checks so operators are reminded if something is not in place, or something is not done properly.'



Complicated

According to Veres some of the AUV systems are complicated to set up and technicians on board a ship could need additionally support to operate them correctly.

Veres plans to control the AUV in the water using multi-agent based control systems, which will pave the way to fully cognitive systems to ensure safe operations.

Several challenges face the team from the start of the project. 'For the modelling phase, we need to decide on which techniques and languages to use and how to test them by formal verification. The modelling must be compatible with the model checking formal verification tools in which the project's computer scientist Dr Alessio Lomuscio is an expert,' said Veres.

Lomuscio, a senior lecturer at Imperial College, will work closely with an engineering systems specialist at Southampton.

Ultimately, more reliable AUV systems could save the industries and scientists who use them a great deal of money.

'AUV projects involve an investment of several million pounds in each vehicle, if you include the labour, maintenance and the staff costs,' said Veres. 'In the end, this methodology could be employed on any number of vehicles. If it saves just one, it's worth it.'