Final Update: Autonomous Ocean Exploration and Conclusions

In my last blog post, I’d like to discuss autonomous ocean-exploring craft and draw some conclusions from the various areas I studied this summer. When one thinks of artificial intelligence’s space applications, autonomous submarines are admittedly not the first things that come to mind. They may yet find use, however, in exploring ocean worlds. For instance, the moon Enceladus has compounds found in its ocean’s plumes that are key to life [1]. The search for life in space may require that we explore ocean worlds. But given their distance from Earth, such exploration would likely be autonomous.

Autonomous ocean exploration, however, has existed for decades – not to explore foreign worlds, but to explore our own. Indeed, exploration of the Earth’s oceans is growing more autonomous. Work such as [2] and [3] develop algorithms to assimilate remotely and locally sensed data for use in autonomous vehicle planning and response. In this sense, the framework for automating water vehicles is similar to that of rovers or satellites. Water vehicles, however, move with more degrees of freedom. In addition, many applications aim to sense chemical or thermal changes rather than visual objects [2], [4]. In the case of [3], researchers develop algorithms to coordinate a group of autonomous vehicles with varying characteristics, assimilating data sensed in situ, on the shore, and by satellites. The researchers deemed such a massive task necessary to understand the marine carbon cycle; exploring a foreign ocean world, though the world may be teeming with questions, might require more simplicity.

To explore the subsurface oceans of Europa or Enceladus, researchers in [4] propose a three-component system. A low-power submersible would explore the ocean autonomously, communicating its data to a base station at the border of ice and water. That base station would have drilled through the ice, setting up a line of communication between it and an orbiting communication relay. This communication relay, likely orbiting the local planet rather than the moon to reduce radiation exposure, would transmit data to Earth. The autonomy component comes in the actions of the submersible, which would search hydrothermal plumes using a spatial nested search, seeking to find locations with maximal plume strength [4]. Unlike Earth-borne autonomous water vehicles, this would have no help from remotely sensed data, nor could it given the below-ice mission. The researchers note that optimization with respect to vehicle power and data capacity is yet to be reached. Indeed, the researchers propose using data summarization methods analogous to those of the Earth Observing-1 mission, discussed in a previous blog post.

This brings us to a final summary of my Monroe research, connecting similar technologies used in disparate missions. General themes connect these missions: the need to autonomously collect data based on human guidelines, the need to respond to collected and remotely sensed data, and the need to reduce the amount of data sent to Earth. These technologies take different implementations based on operating environments and objectives. Course adjustments in orbit, for example, differ in complexity from those made by a (hopefully) land-bound rover. Where possible, I have recommended technologies that I am acquainted with, attempting to address computing constraints as well. My recommendations are, however, not a complete set: they come from a novice in an advanced field. Through this research I have become fascinated with an increasingly important field. The bounds of human knowledge are not set by the fragility of humans; intelligent robotic systems have and will continue to further space exploration, and in doing so, advance science and humanity.



A diagram of the three-part system for exploring Europa (here) or Enceladus [4].


Reference List

[1]        National Aeronautics and Space Administration, “Ocean Worlds,” National Aeronautics and Space Administration. [Online]. Available: [Accessed: Jul. 19, 2018].

[2]        A.F. Thompson, Y. Chao, S. Chien, J. Kinsey, M.M. Flexas, Z.K. Erickson, J. Farrara, D. Fratantoni, A. Branch, S. Chu, M. Troesch, B. Claus, and J. Kepper. “Satellites to Seafloor: Toward Fully Autonomous Ocean Sampling.” Oceanography, vol. 30, no. 2, pp. 160-168, Jun. 2017.

[3]        A.F. Thompson, S. Chien, Y. Chao, and J. Kinsey, “Science-driven Autonomous & Heterogeneous Robotic Networks: A Vision for Future Ocean Observations,” Keck Institute for Space Studies and Caltech Division of Geological and Planetary Sciences, Mar. 29, 2018.

[4]        A. Branch, G. Xu, M.V. Jakuba, C.R. German, S. Chien, J.C. Kinsey, A.D. Bowen, K.P. Hand, and J.S. Seewald, “Autonomous Nested Search for Hydrothermal Venting,” in 2018 ICAPS Workshop on Planning and Robotics, PlanRob 2018, Delft, The Netherlands, June 26, 2018.