We have developed efficient procedures to test many more settings in 2022. Although we have only tested less than 1 percent of these settings as of early February 2022, we already know that moderate improvement is possible, enough to "undo" about 3 years of degradation. However, we have found through in-flight experiments that the settings for channel 0 has a major effect on bit flips in channel 1, and vice-versa, thus there are 33 squared or 1,089 setting to test for each CCD. The field programmable gate arrays (FPGAs) that control the sampling process allow ground input of waveform sampling positions by the ADCs there are a total of 33 valid settings. Fortunately, there is another way to mitigate the problem. ![]() We have mitigated this problem by warming up the focal-plane electronics (FPE) prior to Mars imaging, but in 2019 we reached what was regarded by engineers at Ball Aerospace (who built the camera) as the safe limit for such temperature cycles. This issue was first discovered by taking extra cold images during cruise to Mars, but has steadily worsened over time. Each CCD has two channels (0 and 1), and some of the 26 channels are much worse than others. At relatively cold temperatures, and worsening over time, the ADCs produce useless data because of digitization errors (bit flips). The analog waveforms from the 13 CCD detectors must be converted into digital numbers (DNs) via analog-to-digital converters (ADCs). The HiRISE camera has performed well beyond its design lifetime of 5.4 years, but has an issue that may prove to be life-limiting. Targets for HiRISE are carefully selected to cover exploration priorities (candidate landing sites) and the highest science priorities. Given the high data volume of >70,000 images of ~1 giga-pixel size, only ~4 percent of Mars has been imaged, and much of that is repeat coverage for stereo or change detection, so the unique coverage is from 2-3 percent. With MRO's near-polar orbit and off-nadir pointing, all of Mars' surface can be imaged. ![]() The book is organized in six parts: towards AI transparency methods for interpreting AI systems explaining the decisions of AI systems evaluating interpretability and explanations applications of explainable AI and software for explainable AI.MRO/HiRISE has been imaging Mars since 2006 with pixel scales down to 25-35 cm/pixel for characterization of 1-meter scale features, essential for landers and rovers as well as direct science (>1500 peer-reviewed publications). The 22 chapters included in this book provide a timely snapshot of algorithms, theory, and applications of interpretable and explainable AI and AI techniques that have been proposed recently reflecting the current discourse in this field and providing directions of future development. Explainable AI (XAI) has developed as a subfield of AI, focused on exposing complex AI models to humans in a systematic and interpretable manner. ![]() In pursuit of that objective, ways for humans to verify the agreement between the AI decision structure and their own ground-truth knowledge have been explored. Before deploying an AI system, we see a strong need to validate its behavior, and thus establish guarantees that it will continue to perform as expected when deployed in a real-world environment. For sensitive tasks involving critical infrastructures and affecting human well-being or health, it is crucial to limit the possibility of improper, non-robust and unsafe decisions and actions. A limiting factor for a broader adoption of AI technology is the inherent risks that come with giving up human control and oversight to “intelligent” machines. The development of “intelligent” systems that can take decisions and perform autonomously might lead to faster and more consistent decisions.
0 Comments
Leave a Reply. |
AuthorWrite something about yourself. No need to be fancy, just an overview. ArchivesCategories |