Driving Autonomy: The Critical Role of Operations Research and Artificial Intelligence

Autonomous systems are reshaping the world as we know it in terms of climate management, warehousing, medicine and logistics. The course of history has taught us how technological advancements alter the way economies operate and how people or society coexist. The Industrial Revolution transformed manufacturing and labor. The digital age redefined access to information and communication. Today, we stand at the frontier of another shift led by autonomous systems that are reshaping how we think about logistics, mobility, medicine, climate, infrastructure management and beyond.

From Automation to Autonomy, Why Now?

Automation can be described as an enabler of a system governed by a set of rules that essentially control that system. For example, a thermostat is a good example of automation in which a specific state of home climate is controlled by a specific set of rules. Autonomy, on the other hand, effectively introduces uncertainty into this process, in which the system doesn’t have a clear set of rules but rather has guidelines that help navigate uncertainty while making decisions to overcome new uncertain scenarios. Provided with a map of information, autonomous systems use data and algorithms to make decisions following intelligent reasoning. In other words, autonomous systems operate with a knowledge-based decision-making approach that is more complex than the if-then use cases provided via training. Enabling autonomy involves integrating existing systems with sensors and algorithms. As such, physical autonomous systems are a conglomeration of mechanical systems, sensor technology, connectivity and artificial intelligence (AI). Recently, these systems have been at the forefront of development owing to their promise and value in meeting the sustainable development goals stipulated by the United Nations, which address equitable access to all necessities to make livelihoods safer, more resilient and sustainable for humans [1].

Autonomous Systems at Scale: Enabling Intelligence Across Industries

Autonomous systems operating in fleets, such as connected autonomous vehicles, drones and robots, provide strategic intelligence and support industries and economies in addressing various issues worldwide. Globally, it is estimated that around 60% of companies will adopt these autonomous systems by 2025 [2]. This deployment is expected to improve decision-making processes, safety for human activities and reliability of production. The manufacturing industry is already highly automated and continues to rise.

Beyond manufacturing, fleets of driverless autonomous vehicles are an apt example of how these modern technologies affect our everyday lives. Self-driving cars in the streets of Phoenix, San Francisco and Tokyo are operating as robotaxis providing ride-share services to local citizens [3]. Autonomous vehicles bring benefits in terms of reducing congestion and fuel emissions and improving safety and travel times. However, integrating them in existing public transportation infrastructure requires crucial study and assessment. Similarly, fleets of robots serving a geographical area can be tasked to address various needs such as delivering medicine and food in remote areas, tackling wildfires and monitoring vegetation. Their applications have no limit and are customizable for many tasks. Fleets of specialized robots acting in coordination within enclosed areas like warehouses or other logistical locations carry out tedious tasks of moving packages and organizing inventory in a highly efficient manner. Similarly, robotic arm systems aid in packing packages in an optimized manner to efficiently utilize space in pallets.

Tackling Uncertainty: The Role of O.R. and AI

The question now arises: How do autonomous systems manage the uncertainty to which they are exposed in the real world? Unlike traditional automated systems – working in controlled and predictable worlds – autonomous systems have to decide on the basis of dynamic worlds full of new scenarios, interruptions and incomplete information. Whether it’s an autonomous car changing course in response to an unforeseen roadblock or a delivery drone rerouting because of inclement weather, these systems need more than preprogrammed rules – they need smart structures that enable them to reason, adjust and respond in ambiguity. This is where the fields of operations research (O.R.) and AI play a critical role. They complement each other to offer the mathematical modeling, optimization and learning that allow autonomous systems to make real-time and at-scale decisions, as well as systems to reason in uncertainty, optimize performance and learn from the environment.

Recent advancements in agentic AI reflect the rapidly increasing pace toward autonomous intelligence, which essentially propels the ability to achieve complex, long-horizon objectives independently. These systems are designed to operate with minimal guidance and navigate a vast range of situations without being explicitly programmed [4]. The development of such intelligent agents is the path toward autonomy that cannot merely respond but also plan, learn and improve behavior over time.

Simulation-based approaches are now becoming vital in characterizing emergent behavior in complex multiagent systems. For instance, studies on the resilience of connected and autonomous vehicles under cyber-physical disruptions have demonstrated how agent-based simulation assists in understanding the impact of disruptions on the travel time under vehicle behavior dynamics and redundancy mechanisms [5]. The simulation models facilitate policy analysis, safety evaluation and more resilient strategy design.

In urban transport, O.R. techniques have been applied to deal with uncertainty of demand. For example, researchers consider optimizing fleet size and operation of autonomous and conventional buses. Using a stochastic optimization, the authors demonstrate that adopting autonomous buses enables more flexible routes and lower total cost with service quality preserved in uncertain demand [6]. Similarly, the optimization of chargers for fleets of battery electric buses remains a challenging task. One study suggests an integrated optimization model accounting for charger placement and scheduling under opportunity charging. The outcome is a more efficient system with ensured operational feasibility coupled with reduced total energy and infrastructure costs [7].

In industrial environments such as warehouses, deep reinforcement learning (DRL) has been used to effectively solve high-dimensional decision space problems. A study demonstrates how DRL can be used to create a robust robotic packing system that learns optimal solutions for 3D bin-packing problems – adapting to different object sizes and constraints without hardcoding each possible situation [8]. Navigation under uncertainty is another traditional problem area in which metaheuristic optimization techniques have proven fruitful. The grey wolf optimization algorithm was customized and used for the autonomous robot path-planning problem, in which it assists in finding cost-effective and obstacle-free paths, even in unknown or dynamic environments [9]. Finally, multirobot coordination within logistics networks is facilitated through combined modeling techniques. Storage location assignment and path planning are integrated into one optimization problem, resulting in more throughput and better energy efficiency by coordinating physical layout decisions and routing logic [10].

Together, these studies illustrate the ways in which AI and O.R. enable decision-making in autonomous systems and provide the tools necessary to deal with uncertainty. From transportation and warehousing to planning and system robustness, these techniques are constructing the future of autonomy for the real world. With autonomous systems transitioning across domains, the synergy of O.R. and AI will increasingly determine their effectiveness. From managing real-time fleet management to learning viable policies in random settings, these systems are based on solving core problems related to adaptability, scalability and ethical alignment. Decentralized decision-making increases the flexibility and robustness of autonomous fleets, especially where communication latency, localized disruption or scale renders central control infeasible. Further, it helps distributed agents to make autonomous, context-dependent decisions in pursuit of common or different objectives.

Future work must focus on ensuring the development of algorithms that prioritize safety and transparency, which run efficiently in low-resource settings and are highly resilient to dynamic environments. As autonomy expands into socially important areas – transportation, healthcare, logistics and climate resilience –building transparency, fairness and governance into their design is crucial. In this new world, O.R. and AI are not optimization tools in isolation; they are the drivers of smart, adaptive and decentralized decision-making that can enable reliable, resilient and sustainable systems for the future.

References

1. Guenat, S., Purnell, P., Davies, Z. G., Nawrath, M., Stringer, L. C., Babu, G. R., et al., 2022, “Meeting sustainable development goals via robotics and autonomous systems,” Nature Communications, Vol. 13, No. 1, p. 3559.
2. Brown, S., Hingel, G., Ratcheva, V. & Zahidi, S., 2020, “World Economic Forum: The future of jobs report 2020,” October, World Economic Forum.
3. Waymo, 2024, “Partnering with Nihon Kotsu and GO on our first international road trip,” December 12, Waymo, https://waymo.com/blog/2024/12/partnering-with-nihon-kotsu-and-go-on-our-first-international-road-trip.
4. Acharya, D. B., Kuppan, K. & Divya, B., 2025, “Agentic AI: Autonomous Intelligence for Complex Goals–A Comprehensive Survey,” IEEE Access, Vol. 13, pp. 18912-18936.
5. Sripathanallur Murali, P., Mohebbi, S. & Zaman, M., 2025, “Resilience of connected and autonomous vehicles to cyber-physical disruptions,” SIMULATION, https://doi.org/10.1177/00375497241308575.
6. Tian, Q., Lin, Y. H. & Wang, D. Z., 2021, “Autonomous and conventional bus fleet optimization for fixed-route operations considering demand uncertainty,” Transportation, Vol. 48, No. 5, pp. 2735-2763.
7. Wang, Y., Liao, F. & Lu, C., 2022, “Integrated optimization of charger deployment and fleet scheduling for battery electric buses,” Transportation Research Part D: Transport and Environment, Vol. 109, Art. no. 103382.
8. Xiong, H., Ding, K., Ding, W., Peng, J. & Xu, J., 2023, “Towards reliable robot packing system based on deep reinforcement learning,” Advanced Engineering Informatics, Vol. 57, Art. no. 102028.
9. Kumar, R., Singh, L. & Tiwari, R., 2021, “Path planning for the autonomous robots using modified grey wolf optimization approach,” Journal of Intelligent & Fuzzy Systems, Vol. 40, No. 5, pp. 9453-9470.
10. Cai, J., Li, X., Liang, Y. & Ouyang, S., 2021, “Collaborative optimization of storage location assignment and path planning in robotic mobile fulfillment systems,” Sustainability, Vol. 13, No. 10, Art. no. 5644.