By Super User In Formal Reports
Evolutions in autonomy for mine countermeasures: 2017 in-year progress. Furfaro, Thomas C.; Urso, Giorgio. CMRE-FR-2017-011. May 2019.
This report documents the 2017 in-year developments of CMRE's Collaborative Autonomy for Mine Counter Measures (CA-MCM), of the Autonomous Naval Mine Counter Measures (ANMCM) programme. The effort was focused in two areas: the design and early development of a unifying, generic architecture meant to promote the simplification and decoupling of autonomy components to increase portability and reuse while decreasing maintenance; and the continued development of the Distributed, Decoupled Collaborative Autonomy Framework (D2-CAF), CA-MCM?s multi-agent coordination and task allocation tool. The autonomy framework presents a layered strategy based on the use of actions as the basic unit of interaction between layers, which utilise capabilities, which are resources exposed upwards from lower layers as functionalities available to be used. The framework as of yet doesn?t dictate the actual content of the layers themselves, though a natural usage is to construct successively more abstract, complex functionality in the upper layers from the basic provided from the basest layers. Additionally, a notional perception stack is presented, composed of sensors, drivers, and the associated signal processing network. Again, a normalisation of existing signal processing infrastructure is advocated, primarily consisting of atomisation of core functionalities into standalone modules to increase reuse and reconfigurability. The argument for a centralised world model, as part of a perception stack, is made, as it provides forward-looking utilities for multi-dimensional machine learning implementations, and generally promotes autonomy solutions that may take advantage of onboard long term memory of previous experience. The restructuring and ongoing development of D2-CAF is presented, where in 2017 D2-CAF was upgraded to consider more deployment situations, more underlying messaging mechanisms, and more rapid development of task allocation algorithms. D2- CAF was exercised in the GAMEX17 and SHOEX17 trials of 2017..