{"status":"ok","message-type":"work","message-version":"1.0.0","message":{"indexed":{"date-parts":[[2026,6,27]],"date-time":"2026-06-27T11:25:55Z","timestamp":1782559555190,"version":"3.54.5"},"reference-count":57,"publisher":"MDPI AG","issue":"4","license":[{"start":{"date-parts":[[2021,2,13]],"date-time":"2021-02-13T00:00:00Z","timestamp":1613174400000},"content-version":"vor","delay-in-days":0,"URL":"https:\/\/2.zoppoz.workers.dev:443\/https\/creativecommons.org\/licenses\/by\/4.0\/"}],"funder":[{"DOI":"10.13039\/100010661","name":"Horizon 2020 Framework Programme","doi-asserted-by":"publisher","award":["732204"],"award-info":[{"award-number":["732204"]}],"id":[{"id":"10.13039\/100010661","id-type":"DOI","asserted-by":"publisher"}]},{"name":"Swiss State Secretariat for Education, Research and Innovation","award":["16.0159"],"award-info":[{"award-number":["16.0159"]}]}],"content-domain":{"domain":[],"crossmark-restriction":false},"short-container-title":["Sensors"],"abstract":"<jats:p>Standard-sized autonomous vehicles have rapidly improved thanks to the breakthroughs of deep learning. However, scaling autonomous driving to mini-vehicles poses several challenges due to their limited on-board storage and computing capabilities. Moreover, autonomous systems lack robustness when deployed in dynamic environments where the underlying distribution is different from the distribution learned during training. To address these challenges, we propose a closed-loop learning flow for autonomous driving mini-vehicles that includes the target deployment environment in-the-loop. We leverage a family of compact and high-throughput tinyCNNs to control the mini-vehicle that learn by imitating a computer vision algorithm, i.e., the expert, in the target environment. Thus, the tinyCNNs, having only access to an on-board fast-rate linear camera, gain robustness to lighting conditions and improve over time. Moreover, we introduce an online predictor that can choose between different tinyCNN models at runtime\u2014trading accuracy and latency\u2014which minimises the inference\u2019s energy consumption by up to 3.2\u00d7. Finally, we leverage GAP8, a parallel ultra-low-power RISC-V-based micro-controller unit (MCU), to meet the real-time inference requirements. When running the family of tinyCNNs, our solution running on GAP8 outperforms any other implementation on the STM32L4 and NXP k64f (traditional single-core MCUs), reducing the latency by over 13\u00d7 and the energy consumption by 92%.<\/jats:p>","DOI":"10.3390\/s21041339","type":"journal-article","created":{"date-parts":[[2021,2,14]],"date-time":"2021-02-14T05:54:49Z","timestamp":1613282089000},"page":"1339","update-policy":"https:\/\/2.zoppoz.workers.dev:443\/https\/doi.org\/10.3390\/mdpi_crossmark_policy","source":"Crossref","is-referenced-by-count":50,"title":["Robustifying the Deployment of tinyML Models for Autonomous Mini-Vehicles"],"prefix":"10.3390","volume":"21","author":[{"ORCID":"https:\/\/2.zoppoz.workers.dev:443\/https\/orcid.org\/0000-0003-4350-1617","authenticated-orcid":false,"given":"Miguel","family":"de Prado","sequence":"first","affiliation":[{"name":"He-Arc Ingenierie, HES-SO, 2800 Delemont, Switzerland"},{"name":"Integrated System Lab, ETH Zurich, 8092 Zurich, Switzerland"}],"role":[{"vocabulary":"crossref","role":"author"}]},{"ORCID":"https:\/\/2.zoppoz.workers.dev:443\/https\/orcid.org\/0000-0001-7458-4019","authenticated-orcid":false,"given":"Manuele","family":"Rusci","sequence":"additional","affiliation":[{"name":"DEI, University of Bologna, 40126 Bologna, Italy"}],"role":[{"vocabulary":"crossref","role":"author"}]},{"given":"Alessandro","family":"Capotondi","sequence":"additional","affiliation":[{"name":"Department of Physics, Mathematics and Informatics, University of Modena and Reggio Emilia, 41121 Modena, Italy"}],"role":[{"vocabulary":"crossref","role":"author"}]},{"given":"Romain","family":"Donze","sequence":"additional","affiliation":[{"name":"He-Arc Ingenierie, HES-SO, 2800 Delemont, Switzerland"}],"role":[{"vocabulary":"crossref","role":"author"}]},{"given":"Luca","family":"Benini","sequence":"additional","affiliation":[{"name":"Integrated System Lab, ETH Zurich, 8092 Zurich, Switzerland"},{"name":"DEI, University of Bologna, 40126 Bologna, Italy"}],"role":[{"vocabulary":"crossref","role":"author"}]},{"given":"Nuria","family":"Pazos","sequence":"additional","affiliation":[{"name":"He-Arc Ingenierie, HES-SO, 2800 Delemont, Switzerland"}],"role":[{"vocabulary":"crossref","role":"author"}]}],"member":"1968","published-online":{"date-parts":[[2021,2,13]]},"reference":[{"key":"ref_1","doi-asserted-by":"crossref","first-page":"8357","DOI":"10.1109\/JIOT.2019.2917066","article-title":"A 64mW DNN-based Visual Navigation Engine for Autonomous Nano-Drones","volume":"6","author":"Palossi","year":"2019","journal-title":"IEEE Internet Things J."},{"key":"ref_2","doi-asserted-by":"crossref","first-page":"64270","DOI":"10.1109\/ACCESS.2018.2877890","article-title":"Benchmark analysis of representative deep neural network architectures","volume":"6","author":"Bianco","year":"2018","journal-title":"IEEE Access"},{"key":"ref_3","doi-asserted-by":"crossref","unstructured":"Mutlu, O. (2018, January 10\u201314). Processing data where it makes sense in modern computing systems: Enabling in-memory computation. Proceedings of the 2018 7th Mediterranean Conference on Embedded Computing (MECO), Budva, Montenegro.","DOI":"10.1109\/MECO.2018.8405955"},{"key":"ref_4","unstructured":"(2020, December 30). TinyML. Available online: https:\/\/2.zoppoz.workers.dev:443\/https\/www.tinyml.org\/summit\/."},{"key":"ref_5","unstructured":"Banbury, C.R., Reddi, V.J., Lam, M., Fu, W., Fazel, A., Holleman, J., Huang, X., Hurtado, R., Kanter, D., and Lokhmotov, A. (2020). Benchmarking TinyML Systems: Challenges and Direction. arXiv."},{"key":"ref_6","doi-asserted-by":"crossref","first-page":"451","DOI":"10.1016\/S0031-3203(02)00060-2","article-title":"The global k-means clustering algorithm","volume":"36","author":"Likas","year":"2003","journal-title":"Pattern Recognit."},{"key":"ref_7","doi-asserted-by":"crossref","unstructured":"Kramer, O. (2013). K-nearest neighbors. Dimensionality Reduction with Unsupervised Nearest Neighbors, Springer.","DOI":"10.1007\/978-3-642-38652-7"},{"key":"ref_8","doi-asserted-by":"crossref","unstructured":"Rokach, L., and Maimon, O. (2005). Decision trees. Data Mining and Knowledge Discovery Handbook, Springer.","DOI":"10.1007\/0-387-25465-X_9"},{"key":"ref_9","unstructured":"Kr\u00f6se, B., Krose, B., van der Smagt, P., and Smagt, P. (1993). An introduction to neural networks. J. Comput. Sci., Available online: https:\/\/2.zoppoz.workers.dev:443\/http\/citeseerx.ist.psu.edu\/viewdoc\/summary?doi=10.1.1.18.493."},{"key":"ref_10","unstructured":"(2020, December 30). A Survey on Transformer Models in Machine Learning. Available online: https:\/\/2.zoppoz.workers.dev:443\/https\/hannes-stark.com\/assets\/transformer_survey.pdf."},{"key":"ref_11","doi-asserted-by":"crossref","first-page":"36322","DOI":"10.1109\/ACCESS.2019.2905015","article-title":"Recent progress on generative adversarial networks (GANs): A survey","volume":"7","author":"Pan","year":"2019","journal-title":"IEEE Access"},{"key":"ref_12","unstructured":"(2020, December 30). NXPcup. Available online: https:\/\/2.zoppoz.workers.dev:443\/https\/nxpcup.nxp.com\/."},{"key":"ref_13","unstructured":"(2020, December 30). NXP K64F. Available online: https:\/\/2.zoppoz.workers.dev:443\/https\/www.nxp.com."},{"key":"ref_14","doi-asserted-by":"crossref","unstructured":"Flamand, E., Rossi, D., Conti, F., Loi, I., Pullini, A., Rotenberg, F., and Benini, L. (2018, January 10\u201312). GAP-8: A RISC-V SoC for AI at the Edge of the IoT. Proceedings of the 2018 IEEE 29th International Conference on Application-Specific Systems, Architectures and Processors, Milan, Italy.","DOI":"10.1109\/ASAP.2018.8445101"},{"key":"ref_15","doi-asserted-by":"crossref","first-page":"34","DOI":"10.14569\/IJARAI.2013.020206","article-title":"Comparison of supervised and unsupervised learning algorithms for pattern classification","volume":"2","author":"Sathya","year":"2013","journal-title":"Int. J. Adv. Res. Artif. Intell."},{"key":"ref_16","doi-asserted-by":"crossref","unstructured":"Hastie, T., Tibshirani, R., and Friedman, J. (2009). Overview of supervised learning. The Elements of Statistical Learning, Springer.","DOI":"10.1007\/978-0-387-84858-7"},{"key":"ref_17","doi-asserted-by":"crossref","first-page":"1565","DOI":"10.1038\/nbt1206-1565","article-title":"What is a support vector machine?","volume":"24","author":"Noble","year":"2006","journal-title":"Nat. Biotechnol."},{"key":"ref_18","doi-asserted-by":"crossref","unstructured":"Kingma, D.P., and Welling, M. (2019). An introduction to variational autoencoders. arXiv.","DOI":"10.1561\/9781680836233"},{"key":"ref_19","unstructured":"Sutton, R.S., and Barto, A.G. (2018). Reinforcement Learning: An Introduction, MIT Press."},{"key":"ref_20","doi-asserted-by":"crossref","first-page":"279","DOI":"10.1007\/BF00992698","article-title":"Q-learning","volume":"8","author":"Watkins","year":"1992","journal-title":"Mach. Learn."},{"key":"ref_21","unstructured":"Mnih, V., Badia, A.P., Mirza, M., Graves, A., Lillicrap, T., Harley, T., Silver, D., and Kavukcuoglu, K. (2016). Asynchronous methods for deep reinforcement learning. International Conference on Machine Learning, PMLR."},{"key":"ref_22","unstructured":"(2020, December 30). Introduction to Imitation Learning. Available online: https:\/\/2.zoppoz.workers.dev:443\/https\/blog.statsbot.co\/introduction-to-imitation-learning-32334c3b1e7a."},{"key":"ref_23","unstructured":"(2020, December 30). ICML 2018: Imitation Learning Tutorial. Available online: https:\/\/2.zoppoz.workers.dev:443\/https\/sites.google.com\/view\/icml2018-imitation-learning\/."},{"key":"ref_24","unstructured":"Pomerleau, D.A. (1989). Alvinn: An autonomous land vehicle in a neural network. Advances in Neural Information Processing Systems, Morgan Kaufmann Publishers Inc."},{"key":"ref_25","unstructured":"Bojarski, M., Yeres, P., Choromanska, A., Choromanski, K., Firner, B., Jackel, L., and Muller, U. (2017). Explaining how a deep neural network trained with end-to-end learning steers a car. arXiv."},{"key":"ref_26","doi-asserted-by":"crossref","unstructured":"Koci\u0107, J., Jovi\u010di\u0107, N., and Drndarevi\u0107, V. (2019). An End-to-End Deep Neural Network for Autonomous Driving Designed for Embedded Automotive Platforms. Sensors, 19.","DOI":"10.3390\/s19092064"},{"key":"ref_27","doi-asserted-by":"crossref","unstructured":"Pan, Y., Cheng, C.A., Saigol, K., Lee, K., Yan, X., Theodorou, E., and Boots, B. (2018). Agile autonomous driving using end-to-end deep imitation learning. Robotics: Science and systems. arXiv.","DOI":"10.15607\/RSS.2018.XIV.056"},{"key":"ref_28","doi-asserted-by":"crossref","unstructured":"Taylor, B., Marco, V.S., Wolff, W., Elkhatib, Y., and Wang, Z. (2018). Adaptive selection of deep learning models on embedded systems. arXiv.","DOI":"10.1145\/3211332.3211336"},{"key":"ref_29","doi-asserted-by":"crossref","unstructured":"Jin, X.B., Yu, X.H., Wang, X.Y., Bai, Y.T., Su, T.L., and Kong, J.L. (2020). Deep learning predictor for sustainable precision agriculture based on internet of things system. Sustainability, 12.","DOI":"10.3390\/su12041433"},{"key":"ref_30","doi-asserted-by":"crossref","first-page":"1567","DOI":"10.1007\/s00170-014-6091-1","article-title":"Online detection of run out in microdrilling of tungsten and titanium alloys","volume":"74","author":"Beruvides","year":"2014","journal-title":"Int. J. Adv. Manuf. Technol."},{"key":"ref_31","first-page":"41","article-title":"Applications of support vector machine (SVM) learning in cancer genomics","volume":"15","author":"Huang","year":"2018","journal-title":"Cancer Genom.-Proteom."},{"key":"ref_32","doi-asserted-by":"crossref","unstructured":"Beruvides, G., Juanes, C., Casta\u00f1o, F., and Haber, R.E. (2015, January 22\u201324). A self-learning strategy for artificial cognitive control systems. Proceedings of the 2015 IEEE 13th International Conference on Industrial Informatics (INDIN), Cambridge, UK.","DOI":"10.1109\/INDIN.2015.7281903"},{"key":"ref_33","doi-asserted-by":"crossref","unstructured":"de Prado, M., Mundy, A., Saeed, R., Denna, M., Pazos, N., and Benini, L. (2020). Automated Design Space Exploration for optimised Deployment of DNN on Arm Cortex-A CPUs. IEEE Trans. Comput.-Aided Des. Integr. Circuits and Syst.","DOI":"10.1109\/TCAD.2020.3046568"},{"key":"ref_34","doi-asserted-by":"crossref","unstructured":"Kuutti, S., Bowden, R., Jin, Y., Barber, P., and Fallah, S. (2019). A Survey of Deep Learning Applications to Autonomous Vehicle Control. arXiv.","DOI":"10.1007\/978-3-031-01502-1_2"},{"key":"ref_35","unstructured":"(2020, December 30). Auto Pilot. Available online: https:\/\/2.zoppoz.workers.dev:443\/https\/www.tesla.com\/autopilot."},{"key":"ref_36","unstructured":"(2020, December 30). DeepRacer. Available online: https:\/\/2.zoppoz.workers.dev:443\/https\/aws.amazon.com\/deepracer\/."},{"key":"ref_37","unstructured":"O\u2019Kelly, M., Sukhil, V., Abbas, H., Harkins, J., Kao, C., Pant, Y.V., Mangharam, R., Agarwal, D., Behl, M., and Burgio, P. (2019). F1\/10: An Open-Source Autonomous Cyber-Physical Platform. arXiv."},{"key":"ref_38","unstructured":"(2020, December 30). DonkeyCar. Available online: github.com\/autorope\/donkeycar."},{"key":"ref_39","unstructured":"Dukhan, M., Wu, Y., and Lu, H. (2019, September 12). QNNPACK: Open Source Library for Optimized Mobile Deep Learning. Available online: https:\/\/2.zoppoz.workers.dev:443\/https\/engineering.fb.com\/ml-applications\/qnnpack\/."},{"key":"ref_40","doi-asserted-by":"crossref","unstructured":"Wang, E., Zhang, Q., Shen, B., Zhang, G., Lu, X., Wu, Q., and Wang, Y. (2014). Intel math kernel library. High-Performance Computing on the Intel\u00ae Xeon Phi\u2122, Springer.","DOI":"10.1007\/978-3-319-06486-4"},{"key":"ref_41","unstructured":"Jacob, B. (2017). gemmlowp: A small self-contained low-precision GEMM library. arXiv."},{"key":"ref_42","unstructured":"STMicroelectronics (2019, September 12). X-CUBE-AI. Available online: https:\/\/2.zoppoz.workers.dev:443\/https\/www.st.com\/en\/embedded-software\/x-cube-ai.html."},{"key":"ref_43","unstructured":"Lai, L., Suda, N., and Chandra, V. (2018). Cmsis-nn: Efficient neural network kernels for Arm cortex-m cpus. arXiv."},{"key":"ref_44","unstructured":"Zhang, Y., Suda, N., Lai, L., and Chandra, V. (2017). Hello edge: Keyword spotting on microcontrollers. arXiv."},{"key":"ref_45","unstructured":"Chowdhery, A., Warden, P., Shlens, J., Howard, A., and Rhodes, R. (2019). Visual Wake Words Dataset. arXiv."},{"key":"ref_46","doi-asserted-by":"crossref","unstructured":"Garofalo, A., Rusci, M., Conti, F., Rossi, D., and Benini, L. (2019). PULP-NN: Accelerating Quantized Neural Networks on Parallel Ultra-Low-Power RISC-V Processors. arXiv.","DOI":"10.23919\/DATE48585.2020.9116529"},{"key":"ref_47","unstructured":"(2020, December 30). Continual Learning. Available online: https:\/\/2.zoppoz.workers.dev:443\/https\/medium.com\/@culurciello\/continual-learning-da7995c24bca."},{"key":"ref_48","unstructured":"Lomonaco, V. (2019). Continual Learning with Deep Architectures. [PhD Thesis, ALMA]."},{"key":"ref_49","doi-asserted-by":"crossref","first-page":"56","DOI":"10.1016\/j.neunet.2019.03.010","article-title":"Continuous learning in single-incremental-task scenarios","volume":"116","author":"Maltoni","year":"2019","journal-title":"Neural Netw."},{"key":"ref_50","doi-asserted-by":"crossref","first-page":"2935","DOI":"10.1109\/TPAMI.2017.2773081","article-title":"Learning without forgetting","volume":"40","author":"Li","year":"2017","journal-title":"IEEE Trans. Pattern Anal. Mach. Intell."},{"key":"ref_51","doi-asserted-by":"crossref","unstructured":"Pellegrini, L., Graffieti, G., Lomonaco, V., and Maltoni, D. (2019). Latent replay for real-time continual learning. arXiv.","DOI":"10.1109\/IROS45743.2020.9341460"},{"key":"ref_52","unstructured":"LeCun, Y. (2020, December 30). LeNet-5, Convolutional Neural Networks. Available online: https:\/\/2.zoppoz.workers.dev:443\/http\/yann.lecun.com\/exdb\/lenet."},{"key":"ref_53","doi-asserted-by":"crossref","unstructured":"Rusci, M., Capotondi, A., Conti, F., and Benini, L. (October, January 30). Work-in-progress: Quantized nns as the definitive solution for inference on low-power arm mcus?. Proceedings of the2018 International Conference on Hardware\/Software Codesign and System Synthesis (CODES+ ISSS), Turin, Italy.","DOI":"10.1109\/CODESISSS.2018.8525915"},{"key":"ref_54","unstructured":"Rusci, M., Capotondi, A., and Benini, L. (2019). Memory-Driven Mixed Low Precision Quantization For Enabling Deep Network Inference On Microcontrollers. arXiv."},{"key":"ref_55","doi-asserted-by":"crossref","first-page":"1","DOI":"10.1145\/3403572","article-title":"Bonseyes AI Pipeline\u2014Bringing AI to You","volume":"1","author":"Prado","year":"2020","journal-title":"ACM Trans Internet Things"},{"key":"ref_56","doi-asserted-by":"crossref","unstructured":"Sakr, F., Bellotti, F., Berta, R., and De Gloria, A. (2020). Machine Learning on Mainstream Microcontrollers. Sensors, 20.","DOI":"10.3390\/s20092638"},{"key":"ref_57","unstructured":"(2020, December 30). STMicroelectronics STM32L476xx. Available online: https:\/\/2.zoppoz.workers.dev:443\/https\/www.st.com\/resource\/en\/datasheet\/stm32l476je.pdf."}],"container-title":["Sensors"],"original-title":[],"language":"en","link":[{"URL":"https:\/\/2.zoppoz.workers.dev:443\/https\/www.mdpi.com\/1424-8220\/21\/4\/1339\/pdf","content-type":"unspecified","content-version":"vor","intended-application":"similarity-checking"}],"deposited":{"date-parts":[[2025,10,11]],"date-time":"2025-10-11T05:23:54Z","timestamp":1760160234000},"score":1,"resource":{"primary":{"URL":"https:\/\/2.zoppoz.workers.dev:443\/https\/www.mdpi.com\/1424-8220\/21\/4\/1339"}},"subtitle":[],"short-title":[],"issued":{"date-parts":[[2021,2,13]]},"references-count":57,"journal-issue":{"issue":"4","published-online":{"date-parts":[[2021,2]]}},"alternative-id":["s21041339"],"URL":"https:\/\/2.zoppoz.workers.dev:443\/https\/doi.org\/10.3390\/s21041339","relation":{},"ISSN":["1424-8220"],"issn-type":[{"value":"1424-8220","type":"electronic"}],"subject":[],"published":{"date-parts":[[2021,2,13]]}}}