Adaptive switching optimization for heterogeneous decision modules under operational uncertainty

Heon Kyun Shin; Chaehwa Lee; Jinho Cha; Minyoung Kil; Jaehyuk Choi; Heon Kyun Shin; Chaehwa Lee; Jinho Cha; Minyoung Kil; Jaehyuk Choi

doi:10.3934/jimo.2026144

Journal of Industrial and Management Optimization

2026, Volume 22, Issue 8: 4088-4137. doi: 10.3934/jimo.2026144

Previous Article Next Article

Research article

Adaptive switching optimization for heterogeneous decision modules under operational uncertainty

1.
C4ISR Systems Technology Planning Team, Korea Research Institute for Defense Technology Planning and Advancement (KRIT), Jinju, Republic of Korea
2.
Center for Army Analysis and Simulation, Republic of Korea Army, Gyeryong, Republic of Korea
3.
Department of Computer Science, Gwinnett Technical College, Lawrenceville, GA, USA
4.
C6 (Command, Control, Communications, and Cyber), ROK–U.S. Combined Forces Command, Pyeongtaek, Republic of Korea
5.
Department of AI Convergence Engineering, Gyeongsang National University, Jinju, Republic of Korea
6.
Department of Defense AI and Robotics, Korea National Defense University, Nonsan, Republic of Korea

Received: 16 May 2026 Revised: 16 June 2026 Accepted: 21 July 2026 Published: 28 July 2026
90B70, 90C15, 68T05, 91B32

We developed a stochastic dynamic optimization framework for orchestrating heterogeneous decision modules under operational uncertainty. The modules may differ in quality, latency, reliability, and switching friction. We formulated their orchestration as a finite discounted Markov decision process in which the active module is part of the state and module selection is the control. The reward permits general nondecreasing latency disutility, including linear, threshold-like, power, and convex penalties. We established stationary optimality, switching and continuation regions, pairwise hysteresis, latency-driven preference shifts, an exact adaptive–greedy Bellman decomposition, zero-friction equivalence under action-independent exogenous dynamics, and near-optimality bounds for simple policies. A faster module is preferred when its delay advantage, together with switching and continuation effects, exceeds the quality advantage of a slower module. The framework identifies when dynamic orchestration is valuable and when simpler myopic or threshold policies are adequate.
- adaptive switching,
- heterogeneous decision modules,
- stochastic dynamic programming,
- latency–quality tradeoff,
- hysteresis
Citation: Heon Kyun Shin, Chaehwa Lee, Jinho Cha, Minyoung Kil, Jaehyuk Choi. Adaptive switching optimization for heterogeneous decision modules under operational uncertainty[J]. Journal of Industrial and Management Optimization, 2026, 22(8): 4088-4137. doi: 10.3934/jimo.2026144

Related Papers:

Abstract

We developed a stochastic dynamic optimization framework for orchestrating heterogeneous decision modules under operational uncertainty. The modules may differ in quality, latency, reliability, and switching friction. We formulated their orchestration as a finite discounted Markov decision process in which the active module is part of the state and module selection is the control. The reward permits general nondecreasing latency disutility, including linear, threshold-like, power, and convex penalties. We established stationary optimality, switching and continuation regions, pairwise hysteresis, latency-driven preference shifts, an exact adaptive–greedy Bellman decomposition, zero-friction equivalence under action-independent exogenous dynamics, and near-optimality bounds for simple policies. A faster module is preferred when its delay advantage, together with switching and continuation effects, exceeds the quality advantage of a slower module. The framework identifies when dynamic orchestration is valuable and when simpler myopic or threshold policies are adequate.

References

[1]	D. L. Parnas, On the criteria to be used in decomposing systems into modules, Commun. ACM, 15 (1972), 1053–1058. https://doi.org/10.1145/361598.361623 doi: 10.1145/361598.361623
[2]	M. W. Maier, Architecting principles for systems-of-systems, Syst. Eng., 1 (1998), 267–284. https://doi.org/10.1002/(SICI)1520-6858(1998)1:4<267::AID-SYS3>3.0.CO;2-D doi: 10.1002/(SICI)1520-6858(1998)1:4<267::AID-SYS3>3.0.CO;2-D
[3]	T. Alabama, Office of the deputy director for engineering, Systems Engineering Guidebook, 2021.
[4]	M. Jamshidi, Systems of systems engineering: Principles and applications, CRC Press, 2008. https://doi.org/10.1201/9781420065893
[5]	P. Acheson, C. H. Dagli, Modeling resilience in system of systems architecture, Procedia Comput. Sci., 95 (2016), 111–118. https://doi.org/10.1016/j.procs.2016.09.300 doi: 10.1016/j.procs.2016.09.300
[6]	A. M. Madni, D. Erwin, M. Sievers, Constructing models for systems resilience: Challenges, concepts, and formal methods, Systems, 8 (2020), 3. https://doi.org/10.3390/systems8010003 doi: 10.3390/systems8010003
[7]	B. Green, Y. Chen, The principles and limits of algorithm-in-the-loop decision making, Proc. ACM Hum. Comput. Interact., 3 (2019), 1–24. https://doi.org/10.1145/3359152 doi: 10.1145/3359152
[8]	R. Parasuraman, T. B. Sheridan, C. D. Wickens, A model for types and levels of human interaction with automation, IEEE Trans. Syst. Man Cybern. A, 30 (2000), 286–297. https://doi.org/10.1109/3468.844354 doi: 10.1109/3468.844354
[9]	M. R. Endsley, From here to autonomy: Lessons learned from human–automation research, Hum. Factors, 59 (2017), 5–27. https://doi.org/10.1177/0018720816681350 doi: 10.1177/0018720816681350
[10]	Y. Wang, S. H. Chung, Artificial intelligence in safety-critical systems: A systematic review, Ind. Manag. Data Syst., 122 (2022), 442–470. https://doi.org/10.1108/IMDS-07-2021-0419 doi: 10.1108/IMDS-07-2021-0419
[11]	D. D. Woods, Four concepts for resilience and the implications for the future of resilience engineering, Reliab. Eng. Syst. Saf., 141 (2015), 5–9. https://doi.org/10.1016/j.ress.2015.03.018 doi: 10.1016/j.ress.2015.03.018
[12]	E. Miedema, S. Waschull, C. Emmanouilidis, Towards trustworthy artificial intelligence for decision-making: A lifecycle perspective on knowledge- and data-driven artificial intelligence systems, Comput. Ind., 174 (2026), 104409. https://doi.org/10.1016/j.compind.2025.104409 doi: 10.1016/j.compind.2025.104409
[13]	Z. Bai, J. Li, H. Shen, A robust two-stage edge cloud resource allocation under edge–edge collaboration, Eur. J. Oper. Res. 2026. https://doi.org/10.1016/j.ejor.2026.04.043
[14]	S. Sajwan, B. K. Panigrahi, A. K. Srivastava, Multi-stage operational resilience metric-driven optimal service restoration in DER-rich power distribution systems, IEEE Access, 13 (2025), 95275–95287. https://doi.org/10.1109/ACCESS.2025.3574480 doi: 10.1109/ACCESS.2025.3574480
[15]	R. Dobbe, T. K. Gilbert, Y. Mintz, Hard choices in artificial intelligence, Artif. Intell., 300 (2021), 103555. https://doi.org/10.1016/j.artint.2021.103555 doi: 10.1016/j.artint.2021.103555
[16]	B. Shneiderman, Human-centered artificial intelligence: Reliable, safe & trustworthy, Int. J. Hum. Comput. Interact., 36 (2020), 495–504. https://doi.org/10.1080/10447318.2020.1741118 doi: 10.1080/10447318.2020.1741118
[17]	B. Gyevnár, A. Kasirzadeh, AI safety for everyone, Nat. Mach. Intell., 7 (2025), 531–542. https://doi.org/10.1038/s42256-025-01020-y
[18]	J. Zhang, Z. Shao, L. Zhang, J. Benitez, Exploring users' post-adoption use of generative AI: An attitudinal ambivalence perspective, Decis. Support Syst., 197 (2025), 114521. https://doi.org/10.1016/j.dss.2025.114521 doi: 10.1016/j.dss.2025.114521
[19]	M. An, J. Lin, J. Benitez, Effects of artificial intelligence usage and knowledge-based dynamic capabilities on organizational innovation: A configurational approach, Decis. Support Syst., 200 (2026), 114573. https://doi.org/10.1016/j.dss.2025.114573 doi: 10.1016/j.dss.2025.114573
[20]	M. Yu. Kitaev, R. F. Serfozo, M/M/1 Queues with switching costs and hysteretic optimal control, Oper. Res., 47 (1999), 310–318. https://doi.org/10.1287/opre.47.2.310 doi: 10.1287/opre.47.2.310
[21]	R. Batta, O. Berman, Q. Wang, Balancing staffing and switching costs in a service center with flexible servers, Eur. J. Oper. Res., 177 (2007), 924–938. https://doi.org/10.1016/j.ejor.2006.01.008 doi: 10.1016/j.ejor.2006.01.008
[22]	M. E. Mayorga, K. M. Taaffe, R. Arumugam, Allocating flexible servers in serial systems with switching costs, Ann. Oper. Res., 172 (2009), 231–242. https://doi.org/10.1007/s10479-009-0575-7 doi: 10.1007/s10479-009-0575-7
[23]	Y. Han, Z. Wang, Optimal switching policy for batch servers, Oper. Res. Lett., 51 (2023), 560–567. https://doi.org/10.1016/j.orl.2023.09.004 doi: 10.1016/j.orl.2023.09.004
[24]	A. Farrokh, V. Krishnamurthy, R. Schober, Optimal adaptive modulation and coding with switching costs, IEEE Trans. Commun., 57 (2009), 697–706. https://doi.org/10.1109/TCOMM.2009.03.070115 doi: 10.1109/TCOMM.2009.03.070115
[25]	Y. Dimitrakopoulos, A. Burnetas, The value of service rate flexibility in an M/M/1 1ueue with admission control, IISE Trans., 49 (2017), 603–621. https://doi.org/10.48550/arXiv.1205.4315 doi: 10.48550/arXiv.1205.4315
[26]	E. Furman, A. Diamant, M. Kristal, Customer acquisition and retention: A fluid approach for staffing, Prod. Oper. Manag., 30 (2021), 4236–4257. https://doi.org/10.1111/poms.13520 doi: 10.1111/poms.13520
[27]	F. Pasqualetti, F. Dörfler, F. Bullo, Attack detection and identification in cyber-physical systems, IEEE Trans. Autom. Control, 58 (2013), 2715–2729. https://doi.org/10.1109/TAC.2013.2266831 doi: 10.1109/TAC.2013.2266831
[28]	N. Mollá, C. Heavin, A. Rabasa, Data-driven decision making: New opportunities for DSS in data stream contexts, J. Decis. Syst., 31 (2022), 1–15. https://doi.org/10.1080/12460125.2022.2071404 doi: 10.1080/12460125.2022.2071404
[29]	G. Rinaldi, K. Theodorakos, F. C. Garcia, O. M. Agudelo, B. D. Moor, DSS4EX: A decision support system framework to explore artificial intelligence pipelines with an application in time series forecasting, Expert Syst. Appl., 269 (2025), 126421. https://doi.org/10.1016/j.eswa.2025.126421 doi: 10.1016/j.eswa.2025.126421
[30]	E. Mahmoodi, M. Fathi, Data-driven simulation-based decision support system for resource allocation in Industry 4.0 and smart manufacturing, J. Manuf. Syst., 72 (2024), 287–307. https://doi.org/10.1016/j.jmsy.2023.11.019 doi: 10.1016/j.jmsy.2023.11.019
[31]	G. Musick, T. A. O'Neill, B. G. Schelble, N. J. McNeese, J. B. Henke, What happens when humans believe their teammate is an AI? An investigation into humans teaming with autonomy, Comput. Hum. Behav., 122 (2021), 106852. https://doi.org/10.1016/j.chb.2021.106852 doi: 10.1016/j.chb.2021.106852
[32]	M. L. Puterman, Markov decision processes: Discrete stochastic dynamic programming, John Wiley & Sons, New York, 1994. https://doi.org/10.1002/9780470316887
[33]	D. P. Bertsekas, Dynamic programming and optimal control, Athena Scientific, 2020.
[34]	B. Legros, Late-rejection, a strategy to perform an overflow policy, Eur. J. Oper. Res., 281 (2020), 66–76. https://doi.org/10.1016/j.ejor.2019.08.037 doi: 10.1016/j.ejor.2019.08.037
[35]	X. Zhang, W. C. Cheung, Online allocation of reusable resources in nonstationary environments, Math. Oper. Res. 2025, 1–31. https://doi.org/10.1287/moor.2023.0250
[36]	N. Zhang, X. Li, Z. Xu, An evolutionary reinforcement learning framework for joint work package sizing and scheduling with uncertainties, Eur. J. Oper. Res., 332 (2026), 391–409. https://doi.org/10.1016/j.ejor.2026.01.023 doi: 10.1016/j.ejor.2026.01.023
[37]	M. G. Juárez, A. Giret, V. Botti, Semantic and modular orchestration of AI-driven digital twins for industrial interoperability and optimization, J. Ind. Inf. Integr., 48 (2025), 100959. https://doi.org/10.1016/j.jii.2025.100959 doi: 10.1016/j.jii.2025.100959
[38]	W. Su, M. Gao, X. Gao, X. Zhu, D. Fang, Adaptive decision-making with deep Q-network for heterogeneous unmanned aerial vehicle swarms in dynamic environments, Comput. Electr. Eng., 119 (2024), 109621. https://doi.org/10.1016/j.compeleceng.2024.109621 doi: 10.1016/j.compeleceng.2024.109621
[39]	E. Tabassi, Artificial intelligence risk management framework (AI RMF 1.0), NIST Press, 2023. https://doi.org/10.6028/NIST.AI.100-1
[40]	M. Zinkevich, Online convex programming and generalized infinitesimal gradient ascent, AAAI Press, 2003,928–935. Available from: https://people.eecs.berkeley.edu/brecht/cs294docs/week1/03.Zinkevich.pdf.
[41]	E. Hazan, Introduction to online convex optimization, Found. Trends Optim., 2 (2016), 157–325. https://doi.org/10.1561/2400000013 doi: 10.1561/2400000013
[42]	S. Shalev-Shwartz, Online learning and online convex optimization, Found. Trends Mach. Learn., 4 (2012), 107–194. https://doi.org/10.1561/2200000018 doi: 10.1561/2200000018
[43]	A. Steland, Online detection of changes in moment-based projections: When to retrain deep learners or update portfolios? J. Mach. Learn. Res., 27 (2026), 1–50. Available from: https://jmlr.org/papers/v27/23-0274.html.
[44]	P. Zhao, Y. J. Zhang, L. Zhang, Z. H. Zhou, Adaptivity and non-stationarity: Problem-dependent dynamic regret for online convex optimization, J. Mach. Learn. Res., 25 (2024), 1–52. Available from: https://jmlr.org/papers/v25/21-0748.html.
[45]	O. D. Ovuakporie, K. G. Pillai, C. Wang, Y. Wei, Differential moderating effects of strategic and operational reconfiguration on the relationship between open innovation practices and innovation performance, Res. Policy, 50 (2021), 104146. https://doi.org/10.1016/j.respol.2020.104146 doi: 10.1016/j.respol.2020.104146
[46]	S. Y. Wu, G. Chen, M. J. Shi, J. Alonso-Mora, Decentralized multi-agent trajectory planning in dynamic environments with spatiotemporal occupancy grid maps, IEEE Int. Conf. Robot. Autom., 2024, 7208–7214. https://doi.org/10.1109/ICRA57147.2024.10610670
[47]	J. D. Schierman, M. D. DeVore, N. D. Richards, M. A. Clark, Runtime assurance for autonomous aerospace systems, J. Guid. Control Dyn., 43 (2020), 2205–2217. https://doi.org/10.2514/1.G004862 doi: 10.2514/1.G004862
[48]	B. Clark, D. Patt, H. Schramm, Mosaic warfare: Exploiting artificial intelligence and autonomous systems to implement decision-centric operations, 2020. Available from: https://csbaonline.org/resource/mosaic-warfare-exploiting-artificial-intelligence-and-autonomous-systems-to-implement-decision-centric-operations/.
[49]	Department of defense, summary of the joint all-domain command and control (JADC2) strategy, 2022. Available from: https://ebs.publicnow.com/view/CF7575BE41231B4BBFE5042F0A08BCD7479E1F89.
[50]	Department of defense, Hicks announces delivery of initial CJADC2 capability, DOD News, 2024.
[51]	Department of defense, FY 2024 annual performance report for the DoD strategic management plan for FYs 2022–2026 and FY 2025 annual performance plan, 2025. Available from: https://insidedefense.com/document/dods-fy-24-annual-performance-report-strategic-management-plan.
[52]	H. K. Shin, J. Cha, H. S. Kim, Y. J. Park, A study on the technical characteristics of a reference model for the korean command and control system (KCCS) considering the future warfare environment, J. Korea Acad. Ind. Coop. Soc., 27 (2026), 205–215. https://doi.org/10.5762/KAIS.2026.27.5.205 doi: 10.5762/KAIS.2026.27.5.205
[53]	B. Clark, D. Patt, T. A. Walton, Implementing decision-centric warfare: Elevating command and control to gain an optionality advantage, Hudson Institute, Washington DC, 2021.
[54]	M. C. Horowitz, Artificial intelligence, international competition, and the balance of power, Tex. Natl. Secur. Rev., 1 (2018), 37–57. https://doi.org/10.15781/T2639KP49
[55]	M. Alshiekh, R. Bloem, R. Ehlers, B. Könighofer, S. Niekum, U. Topcu, Safe reinforcement learning via shielding, Proc. AAAI Conf. Artif. Intell., 2018, 2669–2676. https://doi.org/10.1609/aaai.v32i1.11797
[56]	A. D. Ames, S. Coogan, M. Egerstedt, G. Notomista, K. Sreenath, P. Tabuada, Control barrier functions: Theory and applications, Eur. Control Conf., (2019), 3420–3431. https://doi.org/10.23919/ECC.2019.8796030
[57]	S. D. Gu, L. Yang, Y. L. Du, G. Chen, F. Walter, J. Wang, A review of safe reinforcement learning: Methods, theories, and applications, IEEE Trans. Pattern Anal. Mach. Intell., 46 (2024), 11216–11235. https://doi.org/10.1109/TPAMI.2024.3457538 doi: 10.1109/TPAMI.2024.3457538
[58]	E. J. Pinker, R. A. Shumsky, The efficiency–quality trade-Off of cross-trained workers, Manuf. Serv. Oper. Manag., 2 (2000), 32–47. https://doi.org/10.1287/msom.2.1.32.23268 doi: 10.1287/msom.2.1.32.23268

jimo-22-08-144-s001.pdf

Reader Comments

Your name:*

Email:*
© 2026 the Author(s), licensee AIMS Press. This is an open access article distributed under the terms of the Creative Commons Attribution License (http://creativecommons.org/licenses/by/4.0)