Search

Submitted in partial fulfillment of the requirements for the degree of Doctor of Philosophy at the Massachusetts Institute of Technology and the Woods Hole Oceanographic Institution September 2022., Automated information gathering allows exploration of environments where data is limited and gathering observations introduces risk, such as underwater and planetary exploration. Typically, exploration has been performed in service of a query, with a unique algorithm developed for each mission. Yet this approach does not allow scientists to respond to novel questions as they are raised. In this thesis, we develop a single approach for a broad range of adaptive sampling missions with risk and limited prior knowledge. To achieve this, we present contributions in planning adaptive missions in service of queries, and modeling multi-attribute environments. First, we define a query language suitable for specifying diverse goals in adaptive sampling. The language fully encompasses objectives from previous adaptive sampling approaches, and significantly extends the possible range of objectives. We prove that queries expressible in this language are not biased in a way that avoids information. We then describe a Monte Carlo tree search approach to plan for all queries in our language, using sample based objective estimators embedded within tree search. This approach outperforms methods that maximize information about all variables in hydrocarbon seep search and fire escape scenarios. Next, we show how to plan when the policy must bound risk as a function of reward. By solving approximating problems, we guarantee risk bounds on policies with large numbers of actions and continuous observations, ensuring that risks are only taken when justified by reward. Exploration is limited by the quality of the environment model, so we introduce Gaussian process models with directed acyclic structure to improve model accuracy under limited data. The addition of interpretable structure allows qualitative expert knowledge of the environment to be encoded through structure and parameter constraints. Since expert knowledge may be incomplete, we introduce efficient structure learning over stru, The work in this thesis was supported by the Exxon Mobil Corporation as part of the MIT Energy Initiative under the project ‘Autonomous System for Deep Sea Hydrocarbon Detection and Monitoring’, NASA’s PSTAR program under the project ‘Cooperative Exploration with Under-actuated Autonomous Vehicles in Hazardous Environments’, and the Vulcan Machine Learning Center for Impact under the project ‘Machine Learning Based Persistent Autonomous Underwater Scientific Studies’.

© 2020 The MITRE Corporation. All Rights Reserved. While the primary purpose of robotic space exploration systems is to gather scientific data, it is equally important that engineering operations are performed and engineering constraints are respected in order to prolong the mission life and ensure the integrity of the observations taken. However, science and engineering operations are often at odds with each other as attempting to obtain the “best” data may violate engineering operations constraints and place the mission at risk. Historically, mission systems engineering has separated the process of planning for science from engineering operations, with the engineering operations constrained to support the science measurement plan with acceptable risk. This task division leads to multiple design iterations between the science and engineering operations which results in compromised, conservative operations that reduce science return and are more brittle than desired. To overcome these limitations, we present an approach for autonomous mission planning that explicitly models and reasons about the coupling between science and engineering operations, resulting in higher science return, while maintaining acceptable levels of risk. Our approach is to develop an information-driven, risk-bounded plan executive that is capable of producing missions satisfying the goals and constraints expressed in these programs. In this paper, we describe in detail the risk-bounded, information-driven execution problem and lay out the architecture used in our information-directed plan executive ‘Enterprise’. We then show the performance of the current version of Enterprise on two space exploration scenarios. Finally, we conclude with thoughts on future work, including on the design of a proposed information-theoretic language that will allow operators and scientists to specify their objectives in terms of questions about scientific phenomena or the configuration of the space system.

© The Author(s), 2021. This article is distributed under the terms of the Creative Commons Attribution License. The definitive version was published in Vrolijk, P., Summa, L., Ayton, B., Nomikou, P., Huepers, A., Kinnaman, F., Sylva, S., Valentine, D., & Camilli, R. Using a Ladder of Seeps with computer decision processes to explore for and evaluate cold seeps on the Costa Rica active margin. Frontiers in Earth Science, 9, (2021): 601019, https://doi.org/10.3389/feart.2021.601019., Natural seeps occur at the seafloor as loci of fluid flow where the flux of chemical compounds into the ocean supports unique biologic communities and provides access to proxy samples of deep subsurface processes. Cold seeps accomplish this with minimal heat flux. While individual expertize is applied to locate seeps, such knowledge is nowhere consolidated in the literature, nor are there explicit approaches for identifying specific seep types to address discrete scientific questions. Moreover, autonomous exploration for seeps lacks any clear framework for efficient seep identification and classification. To address these shortcomings, we developed a Ladder of Seeps applied within new decision-assistance algorithms (Spock) to assist in seep exploration on the Costa Rica margin during the R/V Falkor 181210 cruise in December, 2018. This Ladder of Seeps [derived from analogous astrobiology criteria proposed by Neveu et al. (2018)] was used to help guide human and computer decision processes for ROV mission planning. The Ladder of Seeps provides a methodical query structure to identify what information is required to confirm a seep either: 1) supports seafloor life under extreme conditions, 2) supports that community with active seepage (possible fluid sample), or 3) taps fluids that reflect deep, subsurface geologic processes, but the top rung may be modified to address other scientific questions. Moreover, this framework allows us to identify higher likelihood seep targets based on existing incomplete or easily acquired data, including MBES (Multi-beam echo sounder) water column data. The Ladder of Seeps framework is based on information about the instruments used to collect seep information (e.g., are seeps detectable by the instrument with little chance of false positives?) and contextual criteria about the environment in which the data are collected (e.g., temporal variability of seep flux). Finally, the assembled data are considered in light of a Last-Resort inte, Support for this research was provided through NASA PSTAR Grant #NNX16AL08G and National Science Foundation Navigating the New Arctic grant #1839063. Use of the R/V Falkor and ROV SuBastian were provided through a grant from the Schmidt Ocean Institute. The AUG Nemesis and the Aurora in-situ mass spectrometer was provided through in-kind support from Teledyne Webb Research and Navistry Corp, respectively.

© 2019, Association for the Advancement of Artificial Intelligence (www.aaai.org). All rights reserved. In autonomous exploration a mobile agent must adapt to new measurements to seek high reward, but disturbances cause a probability of collision that must be traded off against expected reward. This paper considers an autonomous agent tasked with maximizing measurements from a Gaussian Process while subject to unbounded disturbances. We seek an adaptive policy in which the maximum allowed probability of failure is constrained as a function of the expected reward. The policy is found using an extension to Monte Carlo Tree Search (MCTS) which bounds probability of failure. We apply MCTS to a sequence of approximating problems, which allows constraint satisfying actions to be found in an anytime manner. Our innovation lies in defining the approximating problems and replanning strategy such that the probability of failure constraint is guaranteed to be satisfied over the true policy. The approach does not need to plan for all measurements explicitly or constrain planning based only on the measurements that were observed. To the best of our knowledge, our approach is the first to enforce probability of failure constraints in adaptive sampling. Through experiments on real bathymetric data and simulated measurements, we show our algorithm allows an agent to take dangerous actions only when the reward justifies the risk. We then verify through Monte Carlo simulations that failure bounds are satisfied.

Thesis: S.M., Massachusetts Institute of Technology, Department of Aeronautics and Astronautics, 2017., Cataloged from PDF version of thesis., Includes bibliographical references (pages 147-150)., Exploration of new environments is often conducted in search of some phenomenon of interest. Examples include the search for extreme forms of life in the deep ocean or under the ice on Europa, or localizing resource deposits on the ocean floor. Exploration of all these environments is dangerous because of uncertainty in the environment and poorly characterized disturbances that can damage the exploration vehicle. Autonomous vehicles allows exploration in those environments where it is too dangerous or expensive to send a human-operated craft. Autonomous exploration has been well-studied from the perspective of information maximization, but information gathering has not been considered with the intention of localizing specific phenomena, nor has it been considered in environments where exploration can threaten the vehicle. This thesis addresses both challenges by introducing Risk-Bounded Adaptive Search, which maximizes the number of phenomena located while bounding the probability of mission failure by a user-defined threshold. The first innovation of this thesis is the development of a new information measure that focuses on locating instances of a specific phenomenon. Search for phenomena of interest is framed as a discrete space Markov Decision Process that is solved using forward search and receding horizon planning, with a reward function specified as the information gained about unobserved instances of the phenomenon of interest from measurements. Using this reward function, the number of phenomena located is increased compared to maximizing conventional information, as it steers the agent towards locations where phenomena are thought to exist so they are not bypassed when the belief state is high. The second innovation is a method of applying risk bounds as a function of the expected information gain of a policy over a planning horizon, in contrast to a static bound. This 'Performance-Guided Risk Bounding' system allows an MDP policy to be found that is sligh, by Benjamin James Ayton., S.M.