• Dr. Ella Atkins
  • University of Michigan
  • 108 Surge Building
  • 4:00 p.m.
  • Faculty Host: Dr. Craig Woolsey

Aircraft and spacecraft are now able to plan and autonomously execute task-level actions as well as optimizing and precisely following 4-dimensional trajectories. Automation has, however, struggled to effectively manage off-nominal situations, a particularly important capability when such events introduce increased risk of losing a costly vehicle or of causing harm to people or property.  This presentation will describe Dr. Atkins’ long-term research in improving safety and robustness through autonomous contingency management with application to aircraft and spacecraft.

Spacecraft operate in a harsh environment, are costly to launch, and experience unavoidable communication delay and bandwidth constraints. The objective of the presented work is to optimize science goal achievement while identifying and managing encountered faults.  The relative value of science data collection is traded with risk of failure to determine an optimal policy for mission execution. Our major innovation is to incorporate fault information based on the dynamics of the spacecraft and on the internal composition of the spacecraft.  Approximate dynamic programming (ADP) is applied to address computational complexity challenges.

Dr. Atkins will also describe her long-term research in emergency flight management aimed at safely recovering and landing a disabled aircraft.  Her research has focused on envelope-adaptive flight planning and guidance, with a more recent project focusing on flight safety assessment. Results from her research will improve safety in manned and unmanned aircraft, and are beginning to inform risk analysis and mitigation processes relevant for certification of small unmanned aircraft systems (UAS).