The objective of this activity is to make advances in the application of innovative Estimation and Filtering Techniques for Space System Navigation. Most widely used algorithm in current industry for the purpose of Navigation is the Extended Kalman filter (EKF). The inherent shortcomings in using the EKF comes from the fundamental assumptions on the Gaussian distribution of the of random variables, time invariance and ergodicity of the random process that tolerate only mild nonlinearities. In time varying and highly nonlinear cases the EKF based navigation filters introduces large errors in the true posterior mean and covariance leading to poor performance and sometimes even to the divergence of the filter.The objective of this activity is to investigate estimation techniques that can overcome the shortcoming in the EKF assumptions and take into account robustness issues. Candidate techniques as, Particle filtering, Unscented Kalman Filter, Nonlinear Optimal filters with Radial bases, H_inifinity based and variations thereof shall be investigated on their merits. Up to now, techniques such as mixtures of Gaussian, polynomial or particle filtering have been disregarded for space applications due mainly to their high CPU load. With the emergence of new processors, it is worth re-addressing the benefits of such approach for space applications ; approaches such as mix of Gaussians are also known to be more robust to initial uncertainties and non-linearities (both at dynamics and measurement levels) and could potentially overcome some limitations encountered in recent activities such as NPAL/PLGTF(non-linearities anisotropy of measurement errors), or HARVD (initialization of EKF for detection due to very high initial uncertainties on canister position). Further application such as Hybrid Vision based Integrated Navigation applications that cover Space transportation, Rover, Entry Descent Landing, rendezvous systems shall be studied.

The activity shall review current estimation techniques employed in space industry and make a analysis in the face of improvement, new demands and functionalities related to space applications in the view of high accuracy control demands ranging from science missions, exploration and vision based control to space transportation guidance navigation and control applications. The study shall execute the robust and high performance estimation and filtering techniques for accurate state reconstruction for hybrid integrated navigation purposes including vision based hybridisation. Integrated navigation systems are based on the fusion of observations from for example GPS (often noisy), Inertial Measurement Units (IMU), and other available sensors such as from vision and other measurement sources such as radars, lidars, laser altimeters, doppler-velocimeter. Typical absolute navigation problems (based on IMU + vision and/or altimeter and/or lidar) for landing require the "matching" with digital elevation maps (DEM) introducing non-linearities. Therefore it is important first to perform a survey of most important new theoretical developments in estimation and filtering in the view of nonlinearity, time varying effects and uncertainties is to be presented. The various candidate techniques are compared on the basis of metrics that address indicators necessary in the whole design process up to implementation. The survey resulting in candidate algorithms is then mapped onto the hybrid navigation needs for GNC space applications. From the mapping a set of space applications benchmarks covering robust and high performance estimation needs shall be developed. A comparative investigation shall be executed to reveal the impact on the performance, the added implications with respect to implementation and the design process. Such an activity shall also address, as part of the trade-offs, the especially investigation on CPU load of the filters for a future flight implementation.The most promising selected estimation algorithms shall be coded and tested using simple yardstick applications. After maturation a number of hybrid navigation test cases shall be executed for - vision based rover (slower estimation) - vision based entry descent and landing (fast estimation) - a launcher and re-entry vehicle application (fast application)- Rendezvous scenario could be also an interesting test case (slow). - the choice of the reference scenarios (among "slow" and "fast") shall be part of the activity (to be agreed by ESA) Using developed metrics a synthesis is performed on the performance of the various hybrid navigation and estimation benchmarks.