From Determinism to Emergence: Systems Engineering a Step Change in Execution on Mars
Stephen R. Kuhn · 2019
Mars rover operators have traditionally controlled the execution of behaviors using meticulously compiled “master” sequences that enforce deterministic timing and ordering of activities. While the approach is effective at managing risk and complexity, it can be inefficient. Determinism demands conservatism in all things and is mitigated only at the cost of fidelity in ground modeling of flight behavior, an elusive goal. Direct manipulation, as well as inspection by mission operators of the automation layers used to build a schedule, is both feasible and time-consuming. The Mars 2020 rover mission plans to introduce a fundamentally different high-level scheduling and execution approach. Atomic activities are constrained by operators in time, resource usage, and a structure of state dependencies, from largely reusable first principles. The rover flight software will create and re-create a schedule of activities, responding to available energy, actual activity durations and execution status, observed temperatures, and other on-board state. The solution space of possible schedules is prohibitively large and probabilistic. Direct manipulation and inspection of automation are no longer possible on the shortened tactical timelines envisioned for the mission. Schedules are permitted and expected to change substantially as execution burns down the plan. Emergent behavior is accessible depending on the interaction of a given formulation of the plan with actual on-board state. The architecture must anticipate and provide capabilities in the flight software and/or ground tools to help steer emergence. It must leverage the facilities of a highly inherited system to the greatest extent feasible. It must function in that small sliver of computational capacity available in an already diminutive processor, lest it cannibalize from executing activities the efficiency it seeks to gain. This paper discusses the challenges and considerations in systems engineering the first planned autonomous high-level activity planner on Mars.