I am mainly an artificial intelligence researcher with strong interests in biology. My primary research interests are:

  • Artificial intelligence:
    • Context-sensitive reasoning for intelligent agents
    • Intelligent control of autonomous agents, especially autonomous underwater vehicles
    • Deep learning for autonomous agents
    • Multiagent systems
    • Case-based and schema-based reasoning
    • Intelligent assistive technology
    • AI in biology and medicine.
  • Computational ecology:
    • Predator-prey relationships
    • Ecosystem modeling
    • Applications of AI and deep learning to biology and ecology
  • Cognitive science
  • Computer science education

At the current time, I am involved with several active projects:

  • Orca – focuses on intelligence, reactive, context-sensitive control of real-world agents, including autonomous underwater vehicles (AUVs) and land robots, and softbots. This project has been going on since I was at UNH, starting in 1990 or so.

  • Context-mediated behavior (CMB) – focuses on representing and reasoning about contextual knowledge for intelligent agents to allow them to automatically behave appropriately for their context; overlaps with Orca, CoDA, others. Recently, I have begun investigating the use of deep learning in relation to context assessment and context-appropriate behavior.

    This is one of my more active projects over the past twenty years or so, and I am proud to have played a part in helping to form the international community of researchers that center around context research and that are generally identified with the CONTEXT series of conferences (International and Interdisciplinary Conference on Modeling and Using Context). The current conference in the series (CONTEXT-21), of which I am a co-chair, is being planned as an online conference due to the COVID pandemic. I am also on the editorial board of the newly-created journal, Modeling and Using Context.

    Currently, I am focused on the uses of deep learning in context assessment and behaving appropriately for the context, as well as learning in context for neural networks (e.g., following up on some of my - ancient - work in the area.

  • Seabird nesting population survey using deep learning – this work, done in collaboration with Cynthia Loftin and others across campus, applies deep-learning and context-based AI to the problem of identifying nesting seabirds from drone imagery. Currently, M.S. student Alex Revello is working on this, as are two M.S. students from Wildlife Ecology, Logan Klein and Meredith Lewis.

  • CoDA (paused for the moment) – focuses on autonomous organization, operation, and reorganization of real-world multiagent systems (MASs). Its initial domain is controlling autonomous oceanographic sampling networks (AOSNs), which gave it its name: Cooperative Distributed AOSN control, or CoDA. These are groups of AUVs and other instrument platforms that cooperate to return long-term data from an area of interest in the ocean. The technology can also be used for other underwater MASs, such as ones to search for plane wreckage, find the source of pollutants, etc.

    CoDA is an umbrella for several subprojects, including ones focused on:

    • Distributed context-sensitive behavior for MASs
    • Development of cooperation protocols for MAS agents to allow self-organization/reorganization
    • Constraint-based task assignment for MASs
  • Simulation of intelligent agents:
    • Game engine-based simulation for AUVs and multi-AUV systems
    • Multi-fidelity simulation – focuses on developing a simulator for CoDA that can handle high-level, low-resolution through low-level, high-resolution simulations with various parts of the modeled system possibly present in different resolutions. Current version uses CLIPS for high-level simulation, CADCON for low-level simulation; next generation will use Lisa and the game engine-based simulator, above.
    • Simulation of cognitively-impaired humans for assistive technology testing, video games, and “serious games” (e.g., battle simulations). This is primarily the work of one of my PhD students, Chris Wilson.
  • Computational ecology:
    • Predator-prey modeling – small-scale modeling of predator-prey interaction, in particular the important players in early marine fouling community succession, nudibranchs (sea slugs) and hydroids.
    • Estuary modeling – the MEME (Maine Ecosystem Model for Estuaries) project, currently on hiatus, focuses on extending the predator-prey modeling project to the level of an entire estuary to model how early community succession impacts ecosystem structure.
    • CASFish (paused for the moment) – focuses on ecosystem modeling for fisheries management. This is a large, multidisciplinary, multi-organization project that MaineSAIL is involved in. The project is centered primarily in the School of Marine Sciences at UMaine, but also involves researchers from MaineSAIL (R.M. Turner, L. Whitsel), Unity College, the Darling Center, Colby College, and elsewhere.
  • Assistive technology: This relatively new project has two foci, one to develop models of cognitively-impaired individuals for research on cognitive prostheses (mentioned previously) , and the other to develop such aids for cognitively-impaired individuals.

  • Computer science education:
    • K-12: I am interested in teaching computer science to elementary students to attract and retain students (especially girls) to the field. In the past, members of MaineSAIL taught 3rd-5th graders programming using Alice and Scratch (R.M. Turner, E.H. Turner, J. Allen (MS’10), A. L. Conlogue (MS’13)).
    • At the college level, I have a long-term study focusing on how to recruit and retain computer science students, particularly women and minorities, using a rigorous, non-programming introduction to computer science in the first year. Part of this has resulted in the creation of a locally-required course, COS 140: Foundations of Computer Science, for which a textbook is being written.
    • “Computing that Matters” project: This UMaine seed grant-funded project (PI: Harlan Onsrud) involves several faculty of the School in creating a modular introduction to programming course for non-majors whose goal is to attract and retain students, particularly women, to the major. It is an active-learning course that teaches basic Scratch and Python skills, then hones those skills via modules on sensors, GIS, robotics (my module), drones, and others. It is being taught for the first time this Spring (2017) as a section of COS 120.
  • Literate programming for interpreted languages: “Traditional” literate programming tools have significant drawbacks when applied to interpreted languages such as Lisp and Python, since the “tangling” (code generation) phase is highly unnatural in those kinds of languages. Since I mostly work in Lisp, this annoyed me enough to write my own literate programming tool, LP/Lisp. Versions of the programs are available here for others to use.