Prev-Talk | Next-Talk | All-Talks | Talks-Sorted | MOOS-DAWG'22

Talk-14: Composable Autonomies: Results from the Field

Conlan Cesar (2), Laurence Boe (1), Michael DeFilippo (2), Michael Benjamin (2), Andrea Munafo (1)

(1) SeeByte ltd, Edinburgh, UK

(2) MIT, Cambridge MA, USA

To provide a greater number of options when performing missions and to extend the robot capabilities, autonomy architectures must be capable of balancing variable levels of autonomy and coordinate a wide range of robotic behaviours. In the majority of today’s autonomy systems, autonomy architectures are considered in isolation, with little effort put into considering how different autonomies can interact together as a system of systems, providing a gain that goes beyond the sum of its individual parts. To overcome this limitation, the Office of Naval Research (ONR) has funded SeeByte and the Massachusetts Institute of Technology (MIT) to jointly explore the feasibility of composing multiple autonomy architectures that can cohabit on a single platform and across multiple vehicles. This allows the capabilities available within each autonomy to be leveraged concurrently based on mission requirements to produce the more capable unmanned system. This composed autonomy system, that can span across multiple local autonomies and remote autonomies, makes it possible for each robot to match capabilities to mission requirements, while retaining a local ability to make quick decisions when facing unexpected situations.

This contribution will discuss results obtained during the final experimental activity of the project that will be held on the Charles River in July 2022. During the experiment, SeeByte’s Neptune and MIT’s MOOS-IvP will be deployed together on two USVs platforms to perform area surveys while guaranteeing desired levels of safety. The two USVs are very different: one vehicle is fully sensorised to have situational awareness whereas the second one is less capable. For each vehicle, Neptune tackles the higher-level planning and provides feasible routes to avoid known obstacles, replanning as the situation evolves. MOOS-IvP receives and executes the higher-level tasking, while also applying Obstacle Avoidance behaviours to reactively avoid obstacles when needed. When operating in isolation, the less capable USV cannot detect potential dangers on the sea surface, making its path planning hazardous within cluttered environments. The collaboration with the more capable vehicle mitigates this risk and creates a composed autonomy system using up to four distributed autonomies with different levels of capabilities.

Categories:

  • USVs
  • Interoperability