DARPA invites submissions of innovative basic or applied research concepts in the technical domain of AI/automation to support Critical Mineral Assessments (CMAs).
Proposals must address two independent and sequential project phases (a Phase 1 Feasibility Study (base) and a Phase 2 Proof of Concept (option)). The periods of performance for these phases are 12 months for the Phase 1 base effort and 6 months for the Phase 2 option effort. Combined Phase 1 base and Phase 2 option efforts for this AIE Opportunity should not exceed 18 months. The Phase 1 (base) award value is limited to USD$700,000. The Phase 2 (option) award value is limited to USD$300,000.
This solicitation is part of DARPA Artificial Intelligence Exploration Announcement DARPA-PA-22-02.
The CriticalMAAS program seeks research in four technical areas. Proposals in each area should aim to develop scalable methods that fit the scope. Owing to high quality standards for scientific data, human input for curation/validation, correction and/or active learning feedback will be essential in all workflow stages (which map fairly closely to TAs). To accommodate these human-in-the-loop requirements, performers in TA4 will be responsible for creating lightweight interfaces and algorithms that will enable human input at all parts of the workflow. Proposers to TA4 should consider existing tools (particularly open source) and how they can enhance such tools to advance the state-of-the-art in mixed initiative human-computer collaboration approaches and build a practical toolkit targeted to complex data-driven expert workflows like those that produce CMAs.
Technical Areas
TA1 – Extracting geospatial data from maps and documents. – producing high-quality georeferenced vector and tabular data from a huge diversity of USGS maps and documents spanning more than 100 years of data collection.
TA2 – Model extraction from knowledge. – developing computer readable models of mineral deposit types that can be used to classify potential mineral sites
TA3 – Mineral potential mapping exploiting multi-modal fusion – developing methods for probabilistic mineral potential mapping for both ‘brownfield’ (extensions of well-explored sites) and “greenfield” (sites with little previous exploration) resources from the fusion of models and diverse datasets such as geophysics, geochemistry, lithology, remote sensing (eg EMIT8), and historical observations.
TA4: Human-in-the-loop learning (HITL) and mixed-initiative modeling. – design, implementation, and validation of the following:
- Relatively low-code interfaces to support custom expert analysis and input across the workflow, with easy comprehension of outputs and inner-workings of TA1-3 tools
- A modular architecture for integration of the TA1-3 tools in environments that include interactive GIS software products and other such applications that are expected to be beneficial to USGS experts
- Approaches for human feedback (eg active learning) to continuously improve algorithms
Proposals due 6 June 2023.
Supporting documents
AIE Opportunity – CriticalMAAS: DARPA-EA-23-01
AIE Program Announcement:DARPA-EA-23-01-02
