Methodologies for Engineering with Plug-and-Learn Components: Synthesis and Analysis Across Abstraction Layers

Abstract:

Cyber-Physical Systems (CPS) that modify themselves to improve performance or repair damage often recast the modular relationships among system components that enable Verification and Validation (V&V). We focus on in-flight control adaptation of damaged Flapping-Wing Micro Air Vehicles (FW-MAV). Each of our three partner institutions is making a related, but distinct, attack on the problem of encapsulating adaptation into “plug- and-learn” modules and using them to adapt flight control in a way that enables, rather than destroys, V&V capability. The coupled controller, wing motion oscillators, and linkage actuated winged airframe is a self-contained Cyber-Physical system. Our work explores multiple means of exploiting in-system interactions to restore correct control behavior after damage and to diagnose the faults that necessitated any corrections applied. In this poster, we present the major products, to date, of each of our three partner institutions. Included are discussions of inference of wing damage as a side effect of wing gait learning, recovery of control allocation via agent-based learning, and extensions to model checking to enable whole-vehicle health analysis.