Description
The objective of the activity is to design, develop and validate an on-board software relying on Artificial Intelligence that autonomously detect, isolate and recover failures at spacecraft level and that performs resource optimisation at constellation level to recover and improve the overall availability of the service.Targeted Improvements: - Enabling autonomous FDIR management at the level of constellations.- Enabling autonomous performance monitoring and failure prediction.- Reduction of operational outage time by at least one order of magnitude.- Reduction of time required to restore nominal service conditions reduced by at least one order of magnitude (increased availability).Description: Several activities have explored the domain of Artificial Intelligence (AI) andmore specifically Machine Learning on the fault detection problem with very good results, moving from the simple Out-Of-Limit approach with the capability of identifying faults from multiple sources and predicting errors before they impact the system. This allowsfor early warnings to be passed to the operator in advance so that appropriate actions can be taken. However, having the operator in the loop can lead to relatively long period where the service is degraded or lost. Reducing this period to the minimum requires developing techniques to autonomously execute system-level recovery actions on-board when a failure is predicted to occur. When such failure is predicted, the on-board software will have to autonomously take actions to prevent the failure when possible and execute are source optimisation procedure to ensure the continuity of the service at constellation level by transferring failed functions or services to another satellite. Overall, the downtime will be reduced, and the lifespan of individual spacecraft will be extended andthe availability at constellation level will be increased. This activity will design, develop, and test an on-board software that implements the autonomous management of failures covering both use cases: - The platform level to manage failures of a function that can impact the service such as the attitude control system that is degrading the pointing performance and therefore can impact thetelecommunication service. Depending on the degradation level, an autonomous end-of-life procedure (e.g., passivation, de-orbiting,etc.)can be initiated.-The Payload level to manage failures of payload component such as autonomously adapting beamforming to optimise transmissions or transfer the impacted service to another spacecraft of the constellation. AI techniques, like reinforcementlearning or Long-Short-Term-Memory (LSTM), have been studied and successfully applied to on-ground applications and several activities are ongoing in ESA to increase the level of autonomy for the ground segment. AI-based on-board Failure Detection Isolation Recovery (FDIR) algorithms have been investigated (in particular on Attitude Orbit Control System (AOCS) use cases) as part of previous ESA studies, focusing only on the fault detection step, with promising results compared to classical FDIR. This activity will start with the review of spacecraft fault and failure model(s) that impact the service provided by satellites in telecommunication constellations. These models will be an input to the autonomous recovering system. An analysis and evaluation of AI techniques will be performed to select the most relevant one for developing a fully autonomous recovering system that covers all satellite and constellationoperational cases. The on-board software will be developed as well as a simulator supporting its validation through the implementation of use cases. The simulator will cover the functions identified in the fault and failure model(s) at spacecraft and constellation level capturing its dynamic and the evolution of available resources. The testing will be completed on a breadboard representative of anon-board computer processing unit for performance evaluation.Footnote: On Delegation Request activities will only be initiatedon the explicit request of at least one National Delegation.