i i I I ! ! I ! I i I I ! I I I I I I i I I i i i I zyxwvutsrqponmlkjihgfedcbaZYXWVUTSRQPONMLK Ihe Impact ofAuWnomy Technology on Spaceed z -am Amhitecfvret Edward B. Gamblo, Jr., California Irstilute of Twhnology, JeI Propulsion Laboratory Reid Simmons, Sthool of Computer Sden:e, Carnsgie Mallon Universily high-level, clOJed-loopcontrol of spacecraft THE AUTHORS DR4 W ON THElR EXPERl.E,\lCE WITH zyx THE crffers considerable benefits to space-flight AUTONOMY TECHNOLOGY DE;MONSTRATION ON NASA's projects. Those benefits can enable whole new classes of missions; however, they are 1 DEEP SPACE ONE MISSION TO DESCRIBE THE WIDE-RAiiGING not without cost. In this article, we describe both the impact that autonomy technology I EFFECT ACTONOMY WlLL HAYE ON THE DEI/ELOPVENT OF has on spaceCratt softwan and the implica- I tion for the software archicecture that arise from those impacts. Some of the impacts are inherent in the challenging problems gener- implementation for handling the added func- a!ly confronted in the spacecraft domain yet tionality provided by autonomy technologies. are exacerbated by autonomy technology. For instance, in contrast to traditional space-' ' craft software, the computation time of many autonomy technologies cannot be predicted more flexible. Similarly. software that can Ti3 increase reliability and enable new mis- 1 react to unexpected contingencies and oppor- sions. spacecraft 5yscems are moving toward tunities will likely have many branch points, more wtonomous operations. In particular, so brute-force, exhaustive testing is not fea- today's spacecraft designers need autonomy sible. These, and many other factors. com- in cases where long communication time plicate the design, implementation, and vali- delays make command and monitoring oper- dation of autonomy technologies. Overview precisely, su the process scheduler must be ntions by ground-based operaton infeasible. We contendthat a well-conceivedsoftware By analyzing data and making more deci- architecture can have significant positive sions 011 buard. spacecraft can handle con- ilnpact on the developrncnt of autonomy tinycncies quicker and more intelligently, can I technologies and on the ability to integrate takc sstions based on current data, and can 1 them with traditional spacecraft functions, react effectively to ncw opportunities. such 2s at:itude control and telemetry. Spacecraft designers also neednew ap- 1 While attitudes: toward softwarc architec- proaches to spacecraft software design and 1 cures are sornetimcs disrnissivc--"it's just SPACECRAFTSOFTWARE. ' ,!. .:, boxes and arrows"-recent research has laid a rigorous foundation for the field of soft- ware architectures.' The analysis we present here, while not formal, attempts 10 discuss in depth the constraints that the domain of deep- space missionsplace on the design and implementation of autonomy technologies and, similarly, describe how the addition of high levels of autonomy affects the overall software architecture. The term software nrcltit&um enconl- p a w s several different notions. For our yur- poses, we consider only architectural sivle and srructure. Architectural style refers to computational concepts that can be uni.. family applied throughout the system. For instance, one system might be designed in an asynchronuus, publish-subscribe style, while another uses a more synchronous, cliont- server model. An architucturc's structure refers to its decomposition into component