The growing demand for Advanced Air Mobility (AAM) is driving the development of air traffic management systems capable of operating under imperfect information conditions. When surveillance data arrives with noise, delay, partial loss, or even temporary unavailability, maintaining safe separation between aircraft becomes a critical challenge. A promising approach is decentralized conflict resolution through multi-agent reinforcement learning, combining Deep Q-Networks with realistic simulated environments. This article explores how this technology can be integrated into custom software solutions, artificial intelligence, and cloud computing to ensure safe operations.
The core problem is that under degraded surveillance, aircraft must make local decisions with incomplete information. Traditional rule-based or centralized control approaches fail when communication is intermittent or sensors are inaccurate. This is where autonomous agents trained with multi-agent reinforcement learning prove their value: they learn robust policies that exploit behavioral patterns and adapt to heterogeneous dynamics, such as those of small drones and electric vertical takeoff and landing aircraft (eVTOL).
In a recent study (arXiv:2607.20547v1), a Deep Q-Network-based framework was developed for conflict resolution in a structured three-dimensional corridor. Policies are trained separately for each aircraft category using local observations and a 14-action space that includes maintaining course, turning, vertical maneuvering, landing, and speed control. The simulation incorporates observation noise, communication delay, data dropout, wind, actuator and model uncertainty, as well as corridor constraints and energy consumption. Results show that the frequency and duration of loss-of-separation increase with traffic density and separation requirements, although most events are resolved within one second. Under safe conditions, agents maintain their motion about 79% of the time; during conflicts, the most common actions are turning (33%), maintaining motion (29%), speed control (25%), and vertical maneuvers (13%). Six Pareto-optimal configurations reveal trade-offs between safety and corridor capacity.
This type of research has direct implications for companies developing custom applications in the aerospace and mobility sector. Implementing multi-agent AI solutions requires not only robust algorithms but also scalable and secure cloud infrastructure. Q2BSTUDIO, as a technology-specialized company, offers AI services that can integrate these models into real-time simulation and control platforms. Additionally, cybersecurity is critical: communication systems between agents must be protected against attacks that may inject noise or malicious delays. Q2BSTUDIO's cybersecurity solutions ensure the integrity and availability of surveillance data even under adverse conditions.
The cloud plays a crucial role. Training deep networks requires significant computing power, which can be hosted on AWS or Azure through Q2BSTUDIO's cloud services. Once trained, policies can be deployed on edge devices to minimize latency. Likewise, analyzing simulation results—such as conflict frequencies, energy consumption, or capacity trade-offs—benefits from Business Intelligence tools like Power BI. Dashboards allow operators to visualize system performance in real time and make informed decisions about airspace management.
Process automation is another enabler. From generating training scenarios to continuous integration of new policies, custom software automation drastically reduces development times. Q2BSTUDIO has an automation division that can personalize machine learning pipelines for this specific domain.
From a technical perspective, the multi-agent DQN conflict resolution framework is scalable to fleets of hundreds of aircraft. Decentralization avoids bottlenecks and single points of failure. Each agent only needs its local observation—position, speed, heading—and minimal communication with neighbors. This is especially relevant in degraded surveillance environments, where shared information may arrive incomplete or outdated. The study results indicate that even with 20% data loss and delays up to 500 ms, the conflict resolution rate remains above 95%.
For companies looking to implement these solutions, the key lies in combining scientific knowledge with software engineering. It is not enough to have a good algorithm; it must be integrated into flight control systems, certified, and interoperable with regulations such as those of the FAA or EASA. Q2BSTUDIO offers consulting and custom software development to overcome these barriers, as well as AWS/Azure cloud services to scale simulation and production infrastructures.
The future of air mobility lies in resilient autonomous systems. Research on conflict resolution under degraded surveillance with multi-agent RL lays the foundation for a safer sky, even when data fails. Companies that embrace this technology, supported by technology partners like Q2BSTUDIO, will be better positioned to lead the next transportation revolution.
