Speaker
Mr
Antonio Tramtola
(TAS-I)
Description
Landing on Mars, Moon and Asteroids: needed computational power and possible platform architecture. Landing on planets is a task which complexity depends on required accuracy. In the current Exomars project the accuracy is in the order of tens of Km: GNC algorithms, as developed by TAS-I for this type of missions, do not require significant computational power. They can be executed by a LEON processor (in the order of 80 MIPS @ 80Hz). If the accuracy in reaching the landing site increases to few hundreds of meters, the GNC requires additional sensors (camera, lidar) and/or dedicated navigation algorithms (i.e. Guided Entry).These elements increase the computational needs and currently available space qualified computing platforms becomes no more adequate. It is estimated that both atmospheric (Mars) and non-atmospheric (Moon, Asteroid) landing requires much more than 1000MIPs of computational power for Absolute Navigation, Relative navigation and Hazard detection. Among these, the Relative Navigation algorithms look the more demanding and drive the requested platform performances because they have to be executed at the control loop frequency (starting at 2 Hz and up to 10 Hz). In the Absolute Navigation, image processing can be used at lower frequency or even from time to time at specific points in the trajectory. The same applies to Hazard detection and Avoidance that is carried out at a well-defined altitude of the landing trajectory for the final correction. All these aspects have been considered and evaluated in two ESA studies (SAGE and VISNAV-EM-1) lead by TAS-I and also in the STEPS project (internal R&D co-founded by Regione Piemonte) on image processing with the objective of defining and testing a computing platform architecture which can be implemented using space qualified components .Such architecture uses a processor coprocessor configuration in which the coprocessor can be an additional more powerful processor or a FPGA or both FPGA+ Processor. These architectures will be presented pointing out the main features and the achievable performances.
