MURI: Smart Adaptive Reliable Teams for Persistent Surveillance (SMARTS)
The Navy has made significant investments into developing light combat ships (LCS), high altitude long endurance (HALE) vehicles, tactical UAVs, and unmanned underwater vehicles (UUVs) for maritime surveillance. However, despite today’s dynamic battlespace, human operators are consumed by the direct control, monitoring and supervision of autonomous vehicles, constantly “tweaking” the system to obtain the required information. It is necessary to rapidly allocate and re-task different assets to support time-critical intelligence needs for persistent surveillance, andreallocate and re-task assets in response to the detection of threats or changes in missions. To leverage the Navy’s investments in the unmanned vehicle space, it is necessary to make dramatic breakthroughs in decentralized planning and control, perception, autonomy and human interaction. The SMARTS project is a basic research program that will achieve the necessary breakthroughs in computational intelligence. We have assembled a team of experts from artificial intelligence, control theory, formal methods, machine learning, robotics and perception. Our collaborative effort will bridge the currently disparate fields of machine learning and control theory and unify symbolic deliberative planning techniques with algorithms for developing sensor-based, reactive controllers. It will combine formal approaches for task specification and description with natural language-based interfaces for human vehicle interaction. Our efforts will be grounded with a focus application and supporting demonstrations. The focus will be a scenario in which teams of ground, aerial, and underwater autonomous agents interacting with a few human supervisors successfully, carry out complex observation tasks with minimal human intervention. SMARTS will provide a new science base for computational intelligence in decentralized teams of autonomous agents that will be organized in two thrusts:
- The development of intelligent individual agents for interpreting and reasoning about their environments,and
- The development of groups of agents that coordinate with other agents, and with humans, to achieve tasks of larger scope than single agent tasks.
We will tackle simultaneously both thrusts, in the context of complex, uncertain, and changing environments to demonstrate that theoretically-guaranteed behaviors translate into robust, adaptive, and highly-capable robot systems with unprecedented abilities for situation awareness and decision-making. Each thrust is organized around five technical areas which we plan to demonstrate in the context of persistent surveillance in maritime domains: (1) perception; (2) reasoning; (3) learning; (4) control; and (5) human interaction. Each of these areas will map onto a unifying hierarchical architecture for SMARTS. We will demonstrate our advances analytically, in a common simulation testbed, and using hardware testbeds that leverage our existing resources. An advisory group called the Think Tank will be organized to provide external perspective and guidance.
This work is a collaborative Multi-University Research Initiative (MURI) project coordinated by MIT under the leadership of Daniela Rus, with Berkeley, Boston University, and the University of Pennsylvania as participating subcontrators.
BU Principal Investigator: Calin Belta
BU Co-Principal Investigator: Christos Cassandras
Sponsor: Office of Naval Research