PUBLICATIONS

Software Adaptation for an Unmanned Undersea Vehicle

Pfeffer, A.1, Wu, C.1, Fry, G.1, Lu, K.1, Marotta, S.1, Reposa, M.1, Shi, Y.2, Satish Kumar, T.2, Knoblock, C.2, Parker, D.3, Muhammad, I.3, and Novakovic, C.3

IEEE Software, Vol. 36, No. 2 (March/April 2019)

Unmanned undersea vehicles (UUVs) are designed to carry out challenging missions in changing environments. To maximize their effectiveness, these vehicles should adapt to system failures (such as battery loss) and environmental changes (such as a force on the UUV). Since it is expensive to develop UUVs, it is also desirable to increase their lifespan by enabling their software to be able adapt to ecosystem changes such as upgraded sensors.

In our Probabilistic Representation of Intent Commitments to Ensure Software Survival (PRINCESS) project, which is part of the Defense Advanced Research Projects Agency (DARPA) Building Resource Adaptive Software Systems (BRASS) program, we are developing methods to adapt the UUV’s software for all these purposes. Our sensor adaptation accommodates new and upgraded sensors as well as compensates for sensor degradation while the UUV is on a mission. Our control adaptation responds to online system failures and environmental changes in real time; we use probabilistic verification techniques to ensure that these adaptations do not result in software behavior that is dangerous for the UUV.

Our recent work on PRINCESS has involved two scenarios for a REMUS 600 UUV. The first scenario involves degrading a Doppler velocity log sensor used for navigation and a simultaneous perturbation to the environment in the form of a high current. Our adaptation reconstructs an estimate of the sensor signal from other sensors and adjusts the parameters of the navigation system’s Kalman filter to account for the increased noise and environmental perturbation. In the second scenario, the UUV undergoes a catastrophic loss of battery power while on a reconnaissance mission for an object on the ocean floor. Our adaptation reconfigures the UUV’s path planner to generate a path that searches as much of the region as possible while still bringing the UUV home safely without running out of power.

1 Charles River Analytics
2 University of Southern California Information Sciences Institute
3 University of Birmingham

For More Information

To learn more or request a copy of a paper (if available), contact Avi Pfeffer.

(Please include your name, address, organization, and the paper reference. Requests without this information will not be honored.)