{"status":"ok","message-type":"work","message-version":"1.0.0","message":{"indexed":{"date-parts":[[2026,7,21]],"date-time":"2026-07-21T13:08:33Z","timestamp":1784639313679,"version":"3.55.0"},"reference-count":36,"publisher":"Association for Computing Machinery (ACM)","issue":"2","license":[{"start":{"date-parts":[[2025,3,24]],"date-time":"2025-03-24T00:00:00Z","timestamp":1742774400000},"content-version":"vor","delay-in-days":0,"URL":"https:\/\/www.acm.org\/publications\/policies\/copyright_policy#Background"}],"content-domain":{"domain":["dl.acm.org"],"crossmark-restriction":true},"short-container-title":["ACM Trans. Cyber-Phys. Syst."],"published-print":{"date-parts":[[2025,4,30]]},"abstract":"<jats:p>Dual Modular Redundancy (DMR) and Triple Modular Redundancy (TMR), often with some form of diversity, are used in safety-critical systems to realize those functionalities at the highest integrity level providing fault detection and\/or tolerance capabilities. Redundant executions are intended to provide bit-level identical results, and, upon any mismatch, an error is assumed and recovery actions taken as needed. In this article, we note that many emerging AI-based functionalities are intrinsically stochastic (e.g., camera-based object detection), and hence, their correctness must be judged semantically, with room for variations across correct outcomes (e.g., confidence must be above a given threshold, but how much it exceeds the threshold is irrelevant). Building on this observation, we propose strategies to create DMR and TMR implementations of AI-based functionalities that bring not only fault tolerance against random hardware faults but also against AI model inaccuracies. Those strategies, which can be realized with software-only means and ported to virtually any computing platform, build on input data modifications affecting the inference computations, but not the expected semantic output (e.g., introducing some controlled changes in the input data). Moreover, we provide our solution in the form of an open source tool for image and video processing aimed at facilitating the reproducibility of our evaluation results, and enabling others to use it and conduct further research on input transformations.<\/jats:p>","DOI":"10.1145\/3716140","type":"journal-article","created":{"date-parts":[[2025,1,31]],"date-time":"2025-01-31T16:08:11Z","timestamp":1738339691000},"page":"1-26","update-policy":"https:\/\/doi.org\/10.1145\/crossmark-policy","source":"Crossref","is-referenced-by-count":5,"title":["Semantic Diverse DMR and TMR for High-Integrity AI-Based Function Efficiency"],"prefix":"10.1145","volume":"9","author":[{"ORCID":"https:\/\/orcid.org\/0000-0002-0856-1991","authenticated-orcid":false,"given":"Mart\u00ed","family":"Caro","sequence":"first","affiliation":[{"name":"Barcelona Supercomputing Center, Barcelona, Spain and Universitat Polit\u00e8cnica de Catalunya, Barcelona, Spain"}],"role":[{"vocabulary":"crossref","role":"author"}]},{"ORCID":"https:\/\/orcid.org\/0000-0001-8103-391X","authenticated-orcid":false,"given":"Axel","family":"Brando","sequence":"additional","affiliation":[{"name":"Barcelona Supercomputing Center, Barcelona, Spain"}],"role":[{"vocabulary":"crossref","role":"author"}]},{"ORCID":"https:\/\/orcid.org\/0000-0001-7951-4028","authenticated-orcid":false,"given":"Jaume","family":"Abella","sequence":"additional","affiliation":[{"name":"Barcelona Supercomputing Center, Barcelona, Spain"}],"role":[{"vocabulary":"crossref","role":"author"}]}],"member":"320","published-online":{"date-parts":[[2025,3,24]]},"reference":[{"key":"e_1_3_2_2_2","unstructured":"Apollo an Open Autonomous Driving Platform. Retrieved October 2023 from http:\/\/apollo.auto\/"},{"key":"e_1_3_2_3_2","volume-title":"22nd Workshop on Automotive Software and System (SPIN)","author":"Abella J.","year":"2024","unstructured":"J. Abella, I. Agirre, J. Fernandez, L. Belategi, C. Donzella, G. Nicosia, and F. Guerrini. 2024. A tale of machine learning process models. In 22nd Workshop on Automotive Software and System (SPIN). Retrieved from https:\/\/safexplain.eu\/publication\/a-tale-of-machine-learning-process-models-aspice-machine-learning-engineering-mle-vs-safexplain-ai-functional-safety-management-ai-fsm\/"},{"key":"e_1_3_2_4_2","doi-asserted-by":"publisher","DOI":"10.23919\/DATE.2019.8715177"},{"key":"e_1_3_2_5_2","doi-asserted-by":"publisher","DOI":"10.1109\/DFT50435.2020.9250750"},{"key":"e_1_3_2_6_2","doi-asserted-by":"publisher","DOI":"10.1109\/DSN-S50200.2020.00045"},{"key":"e_1_3_2_7_2","unstructured":"Jon Ayerdi Asier Iriarte Pablo Valle Ibai Roman Miren Illarramendi and Aitor Arrieta. 2023. Metamorphic runtime monitoring of autonomous driving systems. arXiv:2310.07414. Retrieved from https:\/\/arxiv.org\/abs\/2310.07414"},{"key":"e_1_3_2_8_2","unstructured":"Alexey Bochkovskiy Chien-Yao Wang and Hong-Yuan Mark Liao. 2020. YOLOv4: Optimal speed and accuracy of object detection. arXiv:2004.10934. Retrieved from https:\/\/arxiv.org\/abs\/2004.10934"},{"key":"e_1_3_2_9_2","doi-asserted-by":"publisher","DOI":"10.1109\/LATW.2019.8704548"},{"key":"e_1_3_2_10_2","doi-asserted-by":"crossref","unstructured":"Mart\u00ed Caro Axel Brando and Jaume Abella. 2024. Software-only semantic diverse redundancy for high-integrity AI-based functionalities. In ERTS2024. Toulouse France. Retrieved from https:\/\/hal.science\/hal-04614881","DOI":"10.1145\/3716140"},{"key":"e_1_3_2_11_2","doi-asserted-by":"crossref","unstructured":"Mart\u00ed Caro Axel Brando and Jaume Abella. 2024. GPU Implementation of Semantic Diverse DMR and TMR for High-Integrity AI-Based Functionalities. Retrieved June 2024 from https:\/\/gitlab.bsc.es\/mcaroroc\/semantic-diversity-dmr-tmr","DOI":"10.1145\/3716140"},{"key":"e_1_3_2_12_2","doi-asserted-by":"publisher","DOI":"10.1109\/COMPSAC57700.2023.00013"},{"key":"e_1_3_2_13_2","doi-asserted-by":"publisher","DOI":"10.1145\/3477314.3507161"},{"key":"e_1_3_2_14_2","doi-asserted-by":"publisher","DOI":"10.1109\/IOLTS.2019.8854431"},{"key":"e_1_3_2_15_2","unstructured":"P. Sun et al. 2020. Scalability in perception for autonomous driving: Waymo open dataset. arXiv:1912.04838. Retrieved from https:\/\/arxiv.org\/abs\/1912.04838"},{"key":"e_1_3_2_16_2","doi-asserted-by":"publisher","DOI":"10.1007\/s11263-009-0275-4"},{"key":"e_1_3_2_17_2","volume-title":"8th Workshop on Critical Automotive Applications: Robustness and Safety (CARS), Held in Conjunction with EDCC 2024","author":"Fernandez J.","year":"2024","unstructured":"J. Fernandez, I. Agirre, J. Perez-Cerrolaza, L. Belategi, A. Adell, C. Donzella, and J. Abella. 2024. AI-FSM: Towards functional safety management for artificial intelligence-based critical systems. In 8th Workshop on Critical Automotive Applications: Robustness and Safety (CARS), Held in Conjunction with EDCC 2024. Retrieved from https:\/\/safexplain.eu\/publication\/ai-fsm-towards-functional-safety-management-for-artificial-intelligence-based-critical-systems\/"},{"key":"e_1_3_2_18_2","doi-asserted-by":"publisher","DOI":"10.1109\/TVLSI.2021.3138491"},{"key":"e_1_3_2_19_2","unstructured":"German Association of the Automotive Industry (VDA)\u2014Quality Management Center (QMC). 2023. Automotive SPICE 4.0. Retrieved from https:\/\/vda-qmc.de\/wp-content\/uploads\/2023\/12\/Automotive-SPICE-PAM-v40.pdf"},{"key":"e_1_3_2_20_2","doi-asserted-by":"publisher","DOI":"10.1109\/34.58871"},{"key":"e_1_3_2_21_2","unstructured":"J\u00f3nathan Heras. 2023. CLoDSA Image Augmentation Library for Object Detection. Retrieved October 2023 https:\/\/github.com\/joheras\/CLoDSA"},{"key":"e_1_3_2_22_2","unstructured":"Infineon. 2023. AURIX Multicore 32-bit Microcontroller Family to Meet Safety and Powertrain Requirements of Upcoming Vehicle Generations. Retrieved October 2023 http:\/\/www.infineon.com\/cms\/en\/about-infineon\/press\/press-releases\/2012\/INFATV201205-040.html"},{"key":"e_1_3_2_23_2","unstructured":"International Standards Organization. 2009. ISO\/DIS 26262. Road Vehicles\u2014Functional Safety. Retrieved from https:\/\/www.iso.org\/publication\/PUB200262.html"},{"key":"e_1_3_2_24_2","unstructured":"International Standards Organization and SAE International. 2021. ISO\/SAE PAS 22736:2021: Taxonomy and Definitions for Terms Related to Driving Automation Systems for On-Road Motor Vehicles. Retrieved from https:\/\/www.iso.org\/standard\/73766.html"},{"key":"e_1_3_2_25_2","doi-asserted-by":"publisher","DOI":"10.1109\/DSN53405.2022.00021"},{"key":"e_1_3_2_26_2","unstructured":"Chih-Yang Li. 2023. YOLOv4 Tensorflow Keras Implementation. Retrieved October 2023 from https:\/\/github.com\/taipingeric\/yolo-v4-tf.keras"},{"key":"e_1_3_2_27_2","doi-asserted-by":"publisher","DOI":"10.1145\/3126908.3126964"},{"key":"e_1_3_2_28_2","unstructured":"Tsung-Yi Lin Michael Maire Serge Belongie Lubomir Bourdev Ross Girshick James Hays Pietro Perona Deva Ramanan C. Lawrence Zitnick and Piotr Doll\u00e1r. 2015. Microsoft COCO: Common objects in context. arXiv:1405.0312. Retrieved from https:\/\/arxiv.org\/abs\/1405.0312"},{"key":"e_1_3_2_29_2","doi-asserted-by":"publisher","DOI":"10.3929\/ethz-b-000407661"},{"key":"e_1_3_2_30_2","unstructured":"Lei Mao. 2024. CUDA GEMM Optimization. Retrieved March 2024 from https:\/\/github.com\/leimao\/CUDA-GEMM-Optimization"},{"key":"e_1_3_2_31_2","doi-asserted-by":"publisher","DOI":"10.1109\/IWSSIP48289.2020.9145130"},{"key":"e_1_3_2_32_2","doi-asserted-by":"publisher","DOI":"10.1145\/3626314"},{"key":"e_1_3_2_33_2","unstructured":"SAFEXPLAIN Consortium. 2023. Space Case Study. Retrieved November 2023 from https:\/\/safexplain.eu\/space-case-study\/"},{"key":"e_1_3_2_34_2","unstructured":"Thilo Strauss Markus Hanselmann Andrej Junginger and Holger Ulmer. 2018. Ensemble methods as a defense to adversarial perturbations against deep neural networks. arXiv:1709.03423. Retrieved from https:\/\/arxiv.org\/abs\/1709.03423"},{"key":"e_1_3_2_35_2","unstructured":"Udacity. 2023. Udacity Self-Driving Car Driving Data. Retrieved October 2023 from https:\/\/github.com\/udacity\/self-driving-car"},{"key":"e_1_3_2_36_2","unstructured":"Fisher Yu Haofeng Chen Xin Wang Wenqi Xian Yingying Chen Fangchen Liu Vashisht Madhavan and Trevor Darrell. 2020. BDD100K: A diverse driving dataset for heterogeneous multitask learning. arXiv:1805.04687. Retrieved from https:\/\/arxiv.org\/abs\/1805.04687"},{"key":"e_1_3_2_37_2","doi-asserted-by":"publisher","DOI":"10.1109\/ICCVW.2019.00120"}],"container-title":["ACM Transactions on Cyber-Physical Systems"],"original-title":[],"language":"en","link":[{"URL":"https:\/\/dl.acm.org\/doi\/10.1145\/3716140","content-type":"unspecified","content-version":"vor","intended-application":"text-mining"},{"URL":"https:\/\/dl.acm.org\/doi\/pdf\/10.1145\/3716140","content-type":"unspecified","content-version":"vor","intended-application":"similarity-checking"}],"deposited":{"date-parts":[[2025,6,19]],"date-time":"2025-06-19T01:18:49Z","timestamp":1750295929000},"score":1,"resource":{"primary":{"URL":"https:\/\/dl.acm.org\/doi\/10.1145\/3716140"}},"subtitle":[],"short-title":[],"issued":{"date-parts":[[2025,3,24]]},"references-count":36,"journal-issue":{"issue":"2","published-print":{"date-parts":[[2025,4,30]]}},"alternative-id":["10.1145\/3716140"],"URL":"https:\/\/doi.org\/10.1145\/3716140","relation":{},"ISSN":["2378-962X","2378-9638"],"issn-type":[{"value":"2378-962X","type":"print"},{"value":"2378-9638","type":"electronic"}],"subject":[],"published":{"date-parts":[[2025,3,24]]},"assertion":[{"value":"2024-06-20","order":0,"name":"received","label":"Received","group":{"name":"publication_history","label":"Publication History"}},{"value":"2025-01-21","order":2,"name":"accepted","label":"Accepted","group":{"name":"publication_history","label":"Publication History"}},{"value":"2025-03-24","order":3,"name":"published","label":"Published","group":{"name":"publication_history","label":"Publication History"}}]}}