Towards the Autonomous Search for Life in the Deep Ocean

The search for life by robotic systems can be approached as a form of foraging behavior similarly characterized by organisms exhibiting evolutionary fitness in the natural world. By evaluating these natural behaviors, we may best inform the behavior of robots in the effort to efficiently detect the existence of biology in remote locations where human exploration is not feasible. Here we present both a robotics system and plan for behavior intended to detect and document life in the deep ocean. This system will further our knowledge of the ocean environment while generating media that can engage the public and inform policy in the realm of marine conservation. Our goal is to facilitate the discovery of new species and habitats through the use of in-situ life detection and adaptive robotic behavior. Through these methods we seek to understand the origins of life on Earth as we help open the door to exploration beyond our own planet. It is our hope that the discovery and documentation of new species and habitats can be leveraged to inform future conservation efforts. Through these efforts we aim to directly contribute to the preservation of our oceans while proving technology suited to accelerate the field of astrobiology. Here we explore the behaviors best suited for a highly efficient autonomous system capable of the search, detection and documentation of life in the deep ocean. Nature has optimized the search for life through the behaviors of foraging and predation. It thus follows that looking to the natural world can best inform the use of multi-sensory information and adaptive behaviors suited for the detection of life. As an example, sharks use certain sensory information such as odor to detect prey from a distance and will leverage additional sensory information as it becomes available with proximity to the prey animal. As a shark approaches its target it will use acoustic noise, turbulence in the water, electric field information and visual ques to acquire its prey. This is radically different from the current methods used to find and inspect targets with robotic systems. Traditional ocean surveys conducted with autonomous marine robots follow a path of pre-programmed rows in which the robot travels along consecutive parallel track lines to cover a box-shaped survey area. After the robot is recovered to a ship, the data is reviewed by human operators who decide the importance of subjects of interest. They then choose sites that should revisited in subsequent missions for high-resolution inspection. This process is thorough but tedious and results in long time intervals between search, detection and inspection. By employing alternative methods of adaptive mission planning based on foraging theory, we can inform behaviors that will enable robotic systems to select areas of investigation based on metrics that are specific to life processes. In the natural world rapid decisions must be made to support evolutionary fitness. This is best achieved by selectively limiting incoming information to reduce the time and energy required to process data. As in nature, the perception of inform to fulfill a biological need is often more advantageous than a full and true perception of reality, so too is the case for a desired utility supported by a robotic system. This model is contrary to standard marine data collection practices but can inform the search for life and minimize the amount of data required to detect, document and classify exciting new discoveries while greatly reducing the need for humans to be part of the decision-making process.