A Quantized State Approach to On-line Simulation for Spacecraft Autonomy

Future space applications will require an increased level of operational autonomy. This calls for declarative methods for spacecraft state estimation and control, so that the spacecraft engineer can focus on modeling the spacecraft rather than implementing all details of the on-line system. Celebrated model based methods such as Kalman filtering techniques and Model Predictive Control (MPC) rely on an on-line model of the system under control that can be simulated in faster than real-time. This becomes a severe challenge when the paradigm of modeling employed is that of hybrid systems where discrete and continuous dynamics co-exists. This paper describes the design and implementation of an efficient engine for simulation of hybrid systems, specifically tailored for on-line applications. The simulation engine, contrary to traditional simulations systems, does not rely on discretisation of time, but instead it works on a discretized state-space where the update of states is determined by a projection of points in time where the trajectory enters a new region. With this approach each state in the model is integrated separately, meaning that sparsity is exploited well. In addition hybrid transitions are located conservatively, i.e. without the need to ever ``roll back'' the simulation in time.