{"status":"ok","message-type":"work","message-version":"1.0.0","message":{"indexed":{"date-parts":[[2026,2,2]],"date-time":"2026-02-02T01:19:21Z","timestamp":1769995161959,"version":"3.49.0"},"reference-count":22,"publisher":"Springer Science and Business Media LLC","issue":"3","license":[{"start":{"date-parts":[[2026,2,1]],"date-time":"2026-02-01T00:00:00Z","timestamp":1769904000000},"content-version":"tdm","delay-in-days":0,"URL":"https:\/\/creativecommons.org\/licenses\/by\/4.0"},{"start":{"date-parts":[[2026,2,1]],"date-time":"2026-02-01T00:00:00Z","timestamp":1769904000000},"content-version":"vor","delay-in-days":0,"URL":"https:\/\/creativecommons.org\/licenses\/by\/4.0"}],"content-domain":{"domain":["link.springer.com"],"crossmark-restriction":false},"short-container-title":["J Supercomput"],"abstract":"Abstract<\/jats:title>\n \n The evolution toward 6G will continue to leverage massive multiple-input multiple-output and millimeter-wave systems, which demand accurate angle-of-arrival (AoA) and angle-of-departure (AoD) estimation. While several deep learning models have demonstrated strong performance for this task, their accuracy, like that of most estimation methods, is often degraded by hardware non-idealities, which can be further exacerbated by time-varying operational factors such as component aging and adverse weather, among others. Building on a pre-trained U-Net architecture with demonstrated competitive performance for AoA\/AoD estimation, we first propose an adaptation mechanism based on fine-tuning with impairment-augmented data. Specifically, we simulate hardware imperfections by introducing random phase errors in the antenna elements, ranging from mild fluctuations to severe signal distortions. The U-Net model with adaptation capabilities is then implemented on an NVIDIA Jetson Orin Nano device, a compact edge platform with heterogeneous computing resources. To this end, we design a co-execution strategy that performs AoA\/AoD estimation (inference) on the CPU while simultaneously fine-tuning the model on the GPU, thus enabling continuous model adaptation to changing environmental or hardware conditions while preserving real-time inference performance. Experimental results show that impairment-aware fine-tuning effectively counters hardware degradation, particularly under significant phase impairments. In such scenarios, the fine-tuned model consistently preserves or even improves estimation accuracy, reducing the Root Mean Square Error (RMSE) by approximately 3.6% and increasing the Probability of Detection (\n \n \n $$P_D$$<\/jats:tex-math>\n \n \n P<\/mml:mi>\n D<\/mml:mi>\n <\/mml:msub>\n <\/mml:math>\n <\/jats:alternatives>\n <\/jats:inline-formula>\n ) by up to 1 percentage point compared to the base model. Furthermore, a detailed energy-performance analysis demonstrates that while maximum frequency settings reduce training time by over 11\n \n \n $$\\times $$<\/jats:tex-math>\n \n \u00d7<\/mml:mo>\n <\/mml:math>\n <\/jats:alternatives>\n <\/jats:inline-formula>\n , they also increase power consumption by more than 5\n \n \n $$\\times $$<\/jats:tex-math>\n \n \u00d7<\/mml:mo>\n <\/mml:math>\n <\/jats:alternatives>\n <\/jats:inline-formula>\n , with optimal energy efficiency achieved at mid-range CPU and high GPU frequencies. This work establishes the feasibility of concurrent training and inference on resource-constrained heterogeneous hardware, paving the way for resilient and autonomous edge intelligence in future 6G systems.\n <\/jats:p>","DOI":"10.1007\/s11227-026-08246-6","type":"journal-article","created":{"date-parts":[[2026,2,1]],"date-time":"2026-02-01T15:10:08Z","timestamp":1769958608000},"update-policy":"https:\/\/doi.org\/10.1007\/springer_crossmark_policy","source":"Crossref","is-referenced-by-count":0,"title":["Energy and robustness trade-offs in adaptive neural mmWave channel estimation on edge devices"],"prefix":"10.1007","volume":"82","author":[{"given":"Eric","family":"Meneses-Albal\u00e1","sequence":"first","affiliation":[],"role":[{"role":"author","vocabulary":"crossref"}]},{"given":"Sa\u00fal","family":"Villaescusa","sequence":"additional","affiliation":[],"role":[{"role":"author","vocabulary":"crossref"}]},{"given":"Jos\u00e9 M.","family":"Bad\u00eda","sequence":"additional","affiliation":[],"role":[{"role":"author","vocabulary":"crossref"}]},{"given":"Germ\u00e1n","family":"Le\u00f3n","sequence":"additional","affiliation":[],"role":[{"role":"author","vocabulary":"crossref"}]},{"given":"Carmen","family":"Botella-Mascarell","sequence":"additional","affiliation":[],"role":[{"role":"author","vocabulary":"crossref"}]},{"given":"Sandra","family":"Roger","sequence":"additional","affiliation":[],"role":[{"role":"author","vocabulary":"crossref"}]}],"member":"297","published-online":{"date-parts":[[2026,2,1]]},"reference":[{"issue":"3","key":"8246_CR1","doi-asserted-by":"publisher","first-page":"55","DOI":"10.1109\/MCOM.001.1900411","volume":"58","author":"M Giordani","year":"2020","unstructured":"Giordani M, Polese M, Mezzavilla M, Rangan S, Zorzi M (2020) Toward 6G networks: use cases and technologies. IEEE Commun Mag 58(3):55\u201361. https:\/\/doi.org\/10.1109\/MCOM.001.1900411","journal-title":"IEEE Commun Mag"},{"issue":"3","key":"8246_CR2","doi-asserted-by":"publisher","first-page":"1478","DOI":"10.1109\/COMST.2024.3364508","volume":"26","author":"Y Zeng","year":"2024","unstructured":"Zeng Y, Chen J, Xu J, Wu D, Xu X, Jin S, Gao X, Gesbert D, Cui S, Zhang R (2024) A tutorial on environment-aware communications via channel knowledge map for 6G. IEEE Commun Surv Tutor 26(3):1478\u20131519. https:\/\/doi.org\/10.1109\/COMST.2024.3364508","journal-title":"IEEE Commun Surv Tutor"},{"issue":"1","key":"8246_CR3","doi-asserted-by":"publisher","first-page":"403","DOI":"10.1109\/TCOMM.2016.2618794","volume":"65","author":"JG Andrews","year":"2017","unstructured":"Andrews JG, Bai T, Kulkarni MN, Alkhateeb A, Gupta AK, Heath RW (2017) Modeling and analyzing millimeter wave cellular systems. IEEE Trans Commun 65(1):403\u2013430. https:\/\/doi.org\/10.1109\/TCOMM.2016.2618794","journal-title":"IEEE Trans Commun"},{"issue":"1","key":"8246_CR4","doi-asserted-by":"publisher","first-page":"173","DOI":"10.1109\/COMST.2018.2869411","volume":"21","author":"M Giordani","year":"2019","unstructured":"Giordani M, Polese M, Roy A, Castor D, Zorzi M (2019) A tutorial on beam management for 3GPP NR at mmWave frequencies. IEEE Commun Surv Tutor 21(1):173\u2013196. https:\/\/doi.org\/10.1109\/COMST.2018.2869411","journal-title":"IEEE Commun Surv Tutor"},{"issue":"9","key":"8246_CR5","doi-asserted-by":"publisher","first-page":"5926","DOI":"10.1109\/TWC.2021.3071315","volume":"20","author":"S Roger","year":"2021","unstructured":"Roger S, Cobos M, Botella-Mascarell C, Fodor G (2021) Fast channel estimation in the transformed spatial domain for analog millimeter wave systems. IEEE Trans Wirel Commun 20(9):5926\u20135941. https:\/\/doi.org\/10.1109\/TWC.2021.3071315","journal-title":"IEEE Trans Wirel Commun"},{"issue":"2","key":"8246_CR6","doi-asserted-by":"publisher","first-page":"93","DOI":"10.1109\/MWC.2019.1800601","volume":"26","author":"Z Qin","year":"2019","unstructured":"Qin Z, Ye H, Li GY, Juang B-HF (2019) Deep learning in physical layer communications. IEEE Wirel Commun 26(2):93\u201399. https:\/\/doi.org\/10.1109\/MWC.2019.1800601","journal-title":"IEEE Wirel Commun"},{"key":"8246_CR7","doi-asserted-by":"publisher","DOI":"10.1109\/TVT.2025.3637908","author":"D Lloria","year":"2025","unstructured":"Lloria D, Roger S, Botella-Mascarell C, Cobos M (2025) Deep-learning-based AoA and AoD estimation in analog millimeter wave MIMO systems. IEEE Trans Veh Technol. https:\/\/doi.org\/10.1109\/TVT.2025.3637908. (To appear in 2026)","journal-title":"IEEE Trans Veh Technol"},{"issue":"1","key":"8246_CR8","doi-asserted-by":"publisher","first-page":"270","DOI":"10.1007\/s11227-024-06795-2","volume":"81","author":"D Lloria","year":"2024","unstructured":"Lloria D, Roger S, Le\u00f3n G, Bad\u00eda JM, Botella-Mascarell C, Belloch JA (2024) Optimizing millimeter wave MIMO channel estimation through GPU-based edge artificial intelligence. J Supercomput 81(1):270. https:\/\/doi.org\/10.1007\/s11227-024-06795-2","journal-title":"J Supercomput"},{"key":"8246_CR9","unstructured":"Bakr O, Johnson M (2009) Impact of phase and amplitude errors on array performance. EECS Department, University of California, Berkeley, Tech. Rep. UCB\/EECS-2009-1, 80"},{"key":"8246_CR10","doi-asserted-by":"publisher","unstructured":"Somjit N, Stemme G, Oberhammer J (2009) Phase error and nonlinearity investigation of millimeter-wave MEMS 7-stage dielectric-block phase shifters. In: 2009 European Microwave Conference, Vol 1\u20133, pp 1872\u20131875. https:\/\/doi.org\/10.1109\/EUMC.2009.5296269","DOI":"10.1109\/EUMC.2009.5296269"},{"key":"8246_CR11","doi-asserted-by":"publisher","DOI":"10.1109\/VTS.2019.8758627","author":"T Moon","year":"2019","unstructured":"Moon T, Gaun J, Hassanieh H (2019) Online millimeter wave phased array calibration based on channel estimation. IEEE. https:\/\/doi.org\/10.1109\/VTS.2019.8758627","journal-title":"IEEE"},{"key":"8246_CR12","doi-asserted-by":"publisher","unstructured":"Dimce S, Amjad MS, Dressler F (2021) mmWave on the road: investigating the weather impact on 60 GHz V2X communication channels. IEEE. https:\/\/doi.org\/10.23919\/WONS51326.2021.9415572","DOI":"10.23919\/WONS51326.2021.9415572"},{"issue":"7","key":"8246_CR13","doi-asserted-by":"publisher","first-page":"880","DOI":"10.1109\/JPROC.2024.3437365","volume":"112","author":"Y Mao","year":"2024","unstructured":"Mao Y, Yu X, Huang K, Angela Zhang Y-J, Zhang J (2024) Green edge AI: a contemporary survey. Proc IEEE 112(7):880\u2013911. https:\/\/doi.org\/10.1109\/JPROC.2024.3437365","journal-title":"Proc IEEE"},{"issue":"12","key":"8246_CR14","doi-asserted-by":"publisher","first-page":"7947","DOI":"10.1109\/TWC.2021.3088910","volume":"20","author":"Q Zeng","year":"2021","unstructured":"Zeng Q, Du Y, Huang K, Leung KK (2021) Energy-efficient resource management for federated edge learning with CPU-GPU heterogeneous computing. IEEE Trans Wirel Commun 20(12):7947\u20137962. https:\/\/doi.org\/10.1109\/TWC.2021.3088910","journal-title":"IEEE Trans Wirel Commun"},{"issue":"5","key":"8246_CR15","doi-asserted-by":"publisher","first-page":"831","DOI":"10.1109\/JSTSP.2014.2334278","volume":"8","author":"A Alkhateeb","year":"2014","unstructured":"Alkhateeb A, El Ayach O, Leus G, Heath RW (2014) Channel estimation and hybrid precoding for millimeter wave cellular systems. IEEE J Sel Top Signal Proc 8(5):831\u2013846. https:\/\/doi.org\/10.1109\/JSTSP.2014.2334278","journal-title":"IEEE J Sel Top Signal Proc"},{"issue":"6","key":"8246_CR16","doi-asserted-by":"publisher","first-page":"1164","DOI":"10.1109\/JSAC.2014.2328154","volume":"32","author":"MR Akdeniz","year":"2014","unstructured":"Akdeniz MR, Liu Y, Samimi MK, Sun S, Rangan S, Rappaport TS, Erkip E (2014) Millimeter wave channel modeling and cellular capacity evaluation. IEEE J Sel Areas Commun 32(6):1164\u20131179. https:\/\/doi.org\/10.1109\/JSAC.2014.2328154","journal-title":"IEEE J Sel Areas Commun"},{"key":"8246_CR17","doi-asserted-by":"publisher","DOI":"10.1109\/ICC.2016.7510902","author":"C Zhang","year":"2016","unstructured":"Zhang C, Guo D, Fan P (2016) Tracking angles of departure and arrival in a mobile millimeter wave channel. IEEE. https:\/\/doi.org\/10.1109\/ICC.2016.7510902","journal-title":"IEEE"},{"issue":"3","key":"8246_CR18","doi-asserted-by":"publisher","first-page":"21","DOI":"10.1109\/MSP.2003.1203207","volume":"20","author":"SC Park","year":"2003","unstructured":"Park SC, Park MK, Kang MG (2003) Super-resolution image reconstruction: a technical overview. IEEE Signal Process Mag 20(3):21\u201336. https:\/\/doi.org\/10.1109\/MSP.2003.1203207","journal-title":"IEEE Signal Process Mag"},{"issue":"10","key":"8246_CR19","doi-asserted-by":"publisher","first-page":"5429","DOI":"10.1109\/TNNLS.2021.3070707","volume":"33","author":"P Pan","year":"2021","unstructured":"Pan P, Zhang Y, Deng Z, Qi W (2021) Deep learning-based 2-D frequency estimation of multiple sinusoidals. IEEE Trans Neural Netw Learn Syst 33(10):5429\u20135440. https:\/\/doi.org\/10.1109\/TNNLS.2021.3070707","journal-title":"IEEE Trans Neural Netw Learn Syst"},{"key":"8246_CR20","doi-asserted-by":"crossref","unstructured":"Lloria D, Roger S, Botella-Mascarell C, Cobos M, Villaescusa S (2025) DL_DOA: deep-learning-based AoA and AoD estimation in analog millimeter wave MIMO systems. GitHub. Accessed: 2025-12-09. https:\/\/github.com\/SandraRoger\/DLDOA","DOI":"10.1145\/3685243.3685247"},{"key":"8246_CR21","unstructured":"NVIDIA: NVIDIA Jetson Linux Developer Guide. Release 32.7.3. https:\/\/docs.nvidia.com\/jetson\/archives\/l4t-archived\/l4t-3273\/ (2022)"},{"key":"8246_CR22","unstructured":"Barrachina S, Barreda M, Catal\u00e1n S, Dolz MF, Fabregat G, Mayo R, Quintana-Ort\u00ed ES (2013) An integrated framework for power-performance analysis of parallel scientific workloads. In: Proceedings of the Third International Conference on Smart Grids, Green Communications and IT Energy-Aware Technologies (ENERGY 2013)"}],"container-title":["The Journal of Supercomputing"],"original-title":[],"language":"en","link":[{"URL":"https:\/\/link.springer.com\/content\/pdf\/10.1007\/s11227-026-08246-6.pdf","content-type":"application\/pdf","content-version":"vor","intended-application":"text-mining"},{"URL":"https:\/\/link.springer.com\/article\/10.1007\/s11227-026-08246-6","content-type":"text\/html","content-version":"vor","intended-application":"text-mining"},{"URL":"https:\/\/link.springer.com\/content\/pdf\/10.1007\/s11227-026-08246-6.pdf","content-type":"application\/pdf","content-version":"vor","intended-application":"similarity-checking"}],"deposited":{"date-parts":[[2026,2,1]],"date-time":"2026-02-01T15:10:09Z","timestamp":1769958609000},"score":1,"resource":{"primary":{"URL":"https:\/\/link.springer.com\/10.1007\/s11227-026-08246-6"}},"subtitle":[],"short-title":[],"issued":{"date-parts":[[2026,2,1]]},"references-count":22,"journal-issue":{"issue":"3","published-online":{"date-parts":[[2026,2]]}},"alternative-id":["8246"],"URL":"https:\/\/doi.org\/10.1007\/s11227-026-08246-6","relation":{},"ISSN":["1573-0484"],"issn-type":[{"value":"1573-0484","type":"electronic"}],"subject":[],"published":{"date-parts":[[2026,2,1]]},"assertion":[{"value":"14 October 2025","order":1,"name":"received","label":"Received","group":{"name":"ArticleHistory","label":"Article History"}},{"value":"14 January 2026","order":2,"name":"accepted","label":"Accepted","group":{"name":"ArticleHistory","label":"Article History"}},{"value":"1 February 2026","order":3,"name":"first_online","label":"First Online","group":{"name":"ArticleHistory","label":"Article History"}},{"order":1,"name":"Ethics","group":{"name":"EthicsHeading","label":"Declarations"}},{"value":"The authors declare no Conflict of interest.","order":2,"name":"Ethics","group":{"name":"EthicsHeading","label":"Conflict of interest"}},{"value":"Not applicable.","order":3,"name":"Ethics","group":{"name":"EthicsHeading","label":"Ethical approval"}}],"article-number":"112"}}