Multi-task learning (MTL) seeks to improve the generalized performance during new testing tasks by exploiting useful information incorporated in related tasks. This paper developed an MTL-based control framework that considers signal temporal logic (STL) specifications applied to air traffic management. MTL settings involve both training and testing stages. In the training stage, an ensemble of tasks is generated by perturbing STL specifications considering the spatiotemporal features, ensuring a diverse yet structured task distribution. Task compliance is measured using the robustness degree, which is computed through STL semantics to quantify how well a state variables trajectory satisfies a set of specifications. Furthermore, robustness degree expressions are inherently non-convex due to log-sum exponential (LSE) terms used to approximate min and max operators, making direct optimization challenging. To address this, we employed sequential convex programming (SCP) with dynamically updated trust regions, which iteratively refine solutions to ensure convergence and stability in the optimization process. In the testing stage, new unseen tasks were introduced, requiring adaptation based on prior knowledge. The optimal solution obtained from the training phase served as a warm start, enabling rapid/few-shot adaptation to new, more perturbed tasks. Experimental validation was performed on two different dynamical systems. The first system corresponds to the spring-mass-damper system. The second system corresponds to a more comprehensive application within the air traffic control problem, scaled down to the context of quad-rotor dynamics. The results demonstrate that the proposed framework successfully satisfies new tasks' specifications within only a few SCP iterations, even under significant perturbations, highlighting the efficiency and adaptability of our approach.
Citation: Andres Arias, Chuangchuang Sun. Multi-task learning for fast online adaptation under signal temporal logic specifications: An air traffic management application[J]. Metascience in Aerospace, 2025, 2(3): 68-88. doi: 10.3934/mina.2025004
Multi-task learning (MTL) seeks to improve the generalized performance during new testing tasks by exploiting useful information incorporated in related tasks. This paper developed an MTL-based control framework that considers signal temporal logic (STL) specifications applied to air traffic management. MTL settings involve both training and testing stages. In the training stage, an ensemble of tasks is generated by perturbing STL specifications considering the spatiotemporal features, ensuring a diverse yet structured task distribution. Task compliance is measured using the robustness degree, which is computed through STL semantics to quantify how well a state variables trajectory satisfies a set of specifications. Furthermore, robustness degree expressions are inherently non-convex due to log-sum exponential (LSE) terms used to approximate min and max operators, making direct optimization challenging. To address this, we employed sequential convex programming (SCP) with dynamically updated trust regions, which iteratively refine solutions to ensure convergence and stability in the optimization process. In the testing stage, new unseen tasks were introduced, requiring adaptation based on prior knowledge. The optimal solution obtained from the training phase served as a warm start, enabling rapid/few-shot adaptation to new, more perturbed tasks. Experimental validation was performed on two different dynamical systems. The first system corresponds to the spring-mass-damper system. The second system corresponds to a more comprehensive application within the air traffic control problem, scaled down to the context of quad-rotor dynamics. The results demonstrate that the proposed framework successfully satisfies new tasks' specifications within only a few SCP iterations, even under significant perturbations, highlighting the efficiency and adaptability of our approach.
| [1] |
Sándor Z (2019) Challenges caused by the unmanned aerial vehicle in the air traffic management. Period Polytech Transp Eng 47: 96–105. https://doi.org/10.3311/PPtr.11204 doi: 10.3311/PPtr.11204
|
| [2] |
Shrestha R, Bajracharya R, Kim S (2021) 6g enabled unmanned aerial vehicle traffic management: A perspective. IEEE Access 9: 91119–91136. https://doi.org/10.1109/ACCESS.2021.3092039 doi: 10.1109/ACCESS.2021.3092039
|
| [3] |
Zhang Y, Yang Q (2021) A survey on multi-task learning. Ieee T Knowl Data Eng 34: 5586–5609. https://doi.org/10.1109/TKDE.2021.3070203 doi: 10.1109/TKDE.2021.3070203
|
| [4] |
Zhang Y, Yang Q (2018) An overview of multi-task learning. Natl Sci Rev 5: 30–43. https://doi.org/10.1093/nsr/nwx105 doi: 10.1093/nsr/nwx105
|
| [5] |
Thung KH, Wee CY (2018) A brief review on multi-task learning. Multimed Tools Appl 77: 29705–29725. https://doi.org/10.1007/s11042-018-6463-x doi: 10.1007/s11042-018-6463-x
|
| [6] | Parnami A, Lee M (2022) Learning from few examples: A summary of approaches to few-shot learning. arXiv preprint. https://doi.org/10.48550/arXiv.2203.04291 |
| [7] | Mao C, Gupta A, Nitin V, et al. (2020) Multitask learning strengthens adversarial robustness. In: Computer Vision–ECCV 2020: 16th European Conference, Glasgow, UK, Proceedings, Part II 16, 158–174. Springer. https://doi.org/10.1007/978-3-030-58536-5_10 |
| [8] |
Caruana R (1997) Multitask learning. Mach learn 28: 41–75. https://doi.org/10.1023/A:1007379606734 doi: 10.1023/A:1007379606734
|
| [9] | Sener O, Koltun V (2018) Multi-task learning as multi-objective optimization. Adv Neur Inform Process Syst 31. |
| [10] | Guo M, Haque A, Huang DA, et al. (2018) Dynamic task prioritization for multitask learning. In: Proceedings of the European conference on computer vision (ECCV), 270–287. |
| [11] | Kendall A, Gal Y, Cipolla R (2018) Multi-task learning using uncertainty to weigh losses for scene geometry and semantics. In: Proceedings of the IEEE conference on computer vision and pattern recognition, 7482–7491. |
| [12] | Vafaeikia P, Namdar K, Khalvati F (2020) A brief review of deep multi-task learning and auxiliary task learning. arXiv preprint. https://doi.org/10.48550/arXiv.2007.01126 |
| [13] |
Gilpin Y, Kurtz V, Lin H (2020) A smooth robustness measure of signal temporal logic for symbolic control. Ieee Contr Syst Lett 5: 241–246. https://doi.org/10.1109/LCSYS.2020.3001875 doi: 10.1109/LCSYS.2020.3001875
|
| [14] | Pant YV, Abbas H, Mangharam R (2017a) Smooth operator: Control using the smooth robustness of temporal logic. In: 2017 IEEE Conference on Control Technology and Applications (CCTA), IEEE, 1235–1240. https://doi.org/10.1109/CCTA.2017.8062628 |
| [15] |
Lindemann L, Dimarogonas DV (2018) Control barrier functions for signal temporal logic tasks. IEEE Contr Syst Lett 3: 96–101. https://doi.org/10.1109/LCSYS.2018.2853182 doi: 10.1109/LCSYS.2018.2853182
|
| [16] |
Lindemann L, Dimarogonas DV (2020) Barrier function based collaborative control of multiple robots under signal temporal logic tasks. Ieee T Control Netw Syst 7: 1916–1928. https://doi.org/10.1109/TCNS.2020.3014602 doi: 10.1109/TCNS.2020.3014602
|
| [17] | Liu W, Xiao W, Belta C (2023) Learning robust and correct controllers from signal temporal logic specifications using barriernet. In: 2023 62nd IEEE Conference on Decision and Control (CDC), 7049–7054. IEEE. https://doi.org/10.1109/CDC49753.2023.10383857 |
| [18] | Yang G, Belta C, Tron R (2020) Continuous-time signal temporal logic planning with control barrier functions. In: 2020 American Control Conference (ACC), 4612–4618. IEEE. https://doi.org/10.23919/ACC45564.2020.9147387 |
| [19] | Xiao W, Belta CA, Cassandras CG (2021) High order control lyapunov-barrier functions for temporal logic specifications. In: 2021 American Control Conference (ACC), 4886–4891. IEEE. https://doi.org/10.23919/ACC50511.2021.9483028 |
| [20] | Madsen C, Vaidyanathan P, Sadraddini S, et al. (2018) Metrics for signal temporal logic formulae. In: 2018 IEEE Conference on Decision and Control (CDC), 1542–1547. IEEE. https://doi.org/10.1109/CDC.2018.8619541 |
| [21] | Mao Y, Acikmese B, Garoche P L, et al. (2022) Successive convexification for optimal control with signal temporal logic specifications. In: Proceedings of the 25th ACM International Conference on Hybrid Systems: Computation and Control, 1–7. https://doi.org/10.1145/3501710.3519518 |
| [22] |
Morgan D, Chung SJ, Hadaegh FY (2014) Model predictive control of swarms of spacecraft using sequential convex programming. J Guid Control Dynam 37: 1725–1740. https://doi.org/10.2514/1.G000218 doi: 10.2514/1.G000218
|
| [23] | Reynolds TP, Mesbahi M (2020) The crawling phenomenon in sequential convex programming. In: 2020 American Control Conference (ACC), 3613–3618. IEEE. https://doi.org/10.23919/ACC45564.2020.9147550 |
| [24] | Hutchins AR, Cummings M, Aubert MC, et al. (2015) Toward the development of a low-altitude air traffic control paradigm for networks of small, autonomous unmanned aerial vehicles. In: AIAA infotech@ aerospace, 1110. https://doi.org/10.2514/6.2015-1110 |
| [25] | Foina AG, Sengupta R, Lerchi P, et al. (2015) Drones in smart cities: Overcoming barriers through air traffic control research. In: 2015 workshop on research, education and development of unmanned aerial systems (RED-UAS), 351–359. IEEE. 10.1109/RED-UAS.2015.7441027 |
| [26] | Pant YV, Abbas H, Mangharam R (2017b) Technical report: Control using the smooth robustness of temporal logic. |
Figures(6)
Andres Arias, Chuangchuang Sun. Multi-task learning for fast online adaptation under signal temporal logic specifications: An air traffic management application[J]. Metascience in Aerospace, 2025, 2(3): 68-88. doi: 10.3934/mina.2025004
DownLoad: