Researchers from the University of Delaware have successfully programmed an AUV to make autonomous decisions during a marine mission
Autonomous underwater vehicles (AUVs) just became a little more autonomous.
A new paper from staff at the University of Delaware’s College of Earth, Ocean and Environment details some exciting improvements in the linking and processing of sensor data, allowing AUVs to adapt their missions according to what they detect.
Mark Moline, director of the university’s School of Marine Science and Policy (as well as co-founder of its Robotic Discovery Laboratory), and Kelly Benoit-Bird of Oregon State University authored the report, published in the Robotics journal. The team formed the initial idea while conducting marine life studies in the Tongue of the Ocean, a deep sea trench in the Bahamas. Their experiment was designed to test whether a modular AUV could be programmed to make decisions autonomously and trigger new missions based on biological information – in their case, the size or concentration of squid – in its environment.
When the sensors detected the correct size and concentration of squid, it triggered a second mission: to report the robot’s position in the water and then run a pre-programmed grid to map the area in finer detail. Their report explains: “While there are numerous autonomous underwater vehicle (AUV) studies demonstrating data feedbacks to inform biological sampling, this study is uncommon in that a processed data product is used to identify a specific target in real time and improve sampling density of that target by a simple autonomous response.”
The writers outfitted a Kongsberg Maritime REMUS600 system with two Simrad EK-60 general purpose transducers/transceivers, one 38 kHz and one 120 kHz (frequencies routinely used in echosounders onboard vessels.) Because of the different ways in which these signals are reflected by marine life (either fish, mammals, or in this case, squid) the return signals can be processed and analysed to indicate what organism is being displayed, and its size.
The AUV’s onboard computers can then use these processed signals to make decisions about what it is scanning – based on measurements of the squid’s size and density – and decide what it should do next. “After the three data products are generated, we determine a threshold for providing a positive signal,” the report notes. “If the criteria set in the data flow for size and number (here 100 squid > 20 cm long) is met, the custom application running on the stack sends a “1” to the vehicle’s RECON (Remote Control) computer, otherwise it sends a zero.”
When a series of “1” signals are detected, the system takes control from the primary navigation computer and pauses its main mission, instead initiating a secondary mission. This sub-mission is undertaken either for a set period, until the mission is completed, or additional sensor data have met another set of programmed conditions. In a mission described in the paper, this was an expanding box grid “beginning with 100m separation between boxes, growing to 200m, and finally one at 500m separation.”
After initial missions offshore California the system was also deployed in the Tongue of the Ocean during July 2015 did. Results were largely positive, with the REMUS successfully detecting squid and initiating its secondary mission to investigate the area further. This, the team said, “showed a number of spatial features in the organism distributions that would have otherwise not been possible by other means,” such as scanning performed from surface vessels.
However, results also suggested that the technology used had some directional limitations: because the REMUS lacks inertial navigation and relies on compass bearings, currents pushed the vehicle slightly off course, resulting in an asymmetrical mission area. That said, correcting such an issue would seem to be straightforward if inertial navigation is factored in.
In a statement, Moline explained: “It was a really simple test that demonstrated that it’s possible to use acoustics to find a species, to have an AUV target specific sizes of that species, and to follow the species, all without having to retrieve and reprogramme the vehicle to hunt for something that will probably be long gone by the time you are ready.”
In a broader context, the paper “demonstrates the powerful combination of multiple data sources, a platform that can improve time/space sampling, real-time data synthesis and autonomous decision-making.” Indeed, the researchers are forthcoming with their analysis of the potential of this decision-making in applications beyond marine life research. In their paper, they comment: “In oil and gas production, there is a need for constant monitoring of the marine environment for both chronic problems and acute effects such as drill cuttings, seeps and leaks. Vehicles could be configured to measure these plumes with fluorometric sensors and internally process data to build a real-time map of a given plume, constantly updating its navigational goals.”
While an accurate example, we at InnovOil think this belies a wealth of further potential. Based on these principles (albeit with some innovation and the right equipment), AUVs could scan pipelines and subsea infrastructure for defects, triggering a secondary mission to capture high-definition images or additional scans if they detect potential cracks or holes. A single mission could then provide engineers with enough detailed information to plan a repair accurately.
Neither is it fantastic to imagine an AUV capable of surveying large areas continuously and comparing data against previous sweeps. If differences are found – e.g. if a pipeline has been unexpectedly moved or been damaged – the AUV’s additional mission could be to investigate further, within defined parameters again taking more detailed scans, images or to map the surrounding area. This could provide valuable information in real-time, without the need for additional or even remote human-controlled surveys. Similar analytical methods could even be applied to aerial vehicles such as drones.
At the moment, these choices are fairly binary. As the researchers note, their system is based on collections of either 1’s or 0’s. But further innovation, as is already occurring across the field of robotics, is likely to mean even greater expansion of AUV capabilities. Moline and Benoit-Bird’s paper may be a small step for man, but it looks like a giant leap for autonomous robots.