Title: An Advanced Architecture for Optimizing Earth Science Data Collection Based Upon Model Predictive Control
Presenting Author: Michael Lieber
Organization: Ball Aerospace & Technologies Corp.
Co-Author(s): Carl Weimer, Reuben Rohrschneider, Lyle Ruppert

Abstract:
The increasing importance of maneuverable and distributed space-based sensor systems has led to exciting developments in large-scale data extraction software and synthesis of complex and enhanced data products. However, beyond spacecraft attitude control system retargeting, the ability to optimize data collection real-time at the sensor level is very limited and constrains the use of platforms with coordinated control and instruments with multiple degrees of freedom. Further work is needed on the lower level, autonomous software to enable fast, optimized control. Two examples of such systems are adaptive lidar and tight formation flying control of future U Class missions. Under NASA Earth Science Technology Office funding, Ball is developing a local, multi-layered control system architecture which communicates with a higher level software layer. The local control formulation is based upon an architecture known as Model Predictive Control (MPC). MPC has found use in many different complex systems where the controlled system is characterized as multivariable, with multiple constraints and possibly nonlinear interactions. These include robotic vision systems, chemical processing, and quad-rotor craft and have been proposed for formation flying spacecraft. MPC optimizes the data collection at each time step from higher level constraints and commands, and is enabled by the increased computational power available in field programmable gate arrays (FPGA) implementations. We discuss development of the MPC architecture for a type of adaptive lidar called Electronically Steerable Flash Lidar (ESFL). ESFL has potentially hundreds of individually steerable laser beamlets and when combined with other sensors pose a large real-time optimization problem well suited to the MPC architecture. The presentation then discusses ways to incorporate an estimator for lidar power return with an evolving scene.