Edge AI enables instant decisions, object classification, and trajectory adjustments without a continuous connection—military and civilian uses, challenges, and prospects.
In summary
The rise of drones with embedded artificial intelligence (Edge AI) is profoundly changing civilian and military applications. Thanks to algorithms capable of processing data locally on the drone—object identification, classification (crop vs. weed, person vs. animal), trajectory adjustment—these systems make decisions in real time without continuous communication with a ground station. This autonomy is a game-changer in agriculture, inspections, surveillance, logistics—and especially in military operations where environments can be degraded or without connectivity. Companies such as Shield AI, Skydio, Edge Autonomy, and Palladyne AI are investing heavily in these capabilities. Major powers are funding these technologies, anticipating a war where speed of decision-making, autonomy, and coordination of UAV fleets will determine the advantage. Issues of ethics, sovereignty, interoperability, and resistance to jamming will need to be monitored.
The concept of embedded drone intelligence
The main concept is based on the embedded processing algorithm (Edge AI) — processing data directly on the drone — which avoids the latency of transferring data to a computing center and reduces dependence on links. This approach enables object classification (e.g., crop vs. weed, person vs. animal), autonomous decision-making (e.g., trajectory adjustment based on a detected obstacle), and reduced communication with a ground station, which improves resilience to jamming or communication failures.
Technically, a drone equipped with a specialized processor (Edge chip) receives sensor feeds (4K camera, LiDAR, infrared). It runs a neural network (e.g., YOLO, ResNet) for real-time detection—latency < 20 ms in some cases. For example, NTT has announced an embedded 4K video processing system capable of operating at 20 W at an altitude of 150 m. Once the object is recognized, an embedded logic module (often using reinforcement learning or rules) determines the course of action: follow the object, avoid it, navigate to a new waypoint, or trigger a sensor. The path is adjusted without using the ground link for seconds or even minutes. This autonomy makes it possible to carry out missions in degraded environments, where satellite or radio links are cut or jammed.
Another part of the system is the coordination of UAV fleets (swarms): each drone shares little information but acts according to common rules and multi-agent learning algorithms (MARL). A study showed a 53% reduction in latency and 63% reduction in energy consumption in a network of five drones compared to a traditional mesh network, and zero collisions in a test with ten drones. This capability paves the way for complex missions without centralized supervision.
Operation and technical architectures
The logical architecture is overlaid with a hardware architecture. The drone carries:
- sensors: visible camera, infrared, LiDAR, compact radar, acoustic microphones, and sometimes electromagnetic sensors.
- an Edge AI computing module: FPGA or ASIC chip, more often an embedded GPU or multi-core CPU designed for real-time AI.
- actuators and flight control systems: to adjust trajectory, speed, and altitude.
- Embedded software: perception pipeline (detection, classification), reasoning module (mission logic, learning), action module (navigation, avoidance).
- Communication interfaces: radio, datalink, sometimes satellite, but these may be in the background.
The typical flow is: movement → sensor capture → classification and decision → flight adjustment. For example, in an agricultural flight, the drone identifies a patch of weeds, replans a route to fly over the target area, triggers a spraying system, and then returns to its normal trajectory. In a military context, the system can detect an enemy vehicle, decide to monitor it, adjust its trajectory to remain hidden, and then trigger a sensor or transmit an alert.
The key is the OODA loop (Observe-Orient-Decide-Act), accelerated by embedded intelligence: “the side that thinks and acts fastest will win.” It is precisely this speed that makes the difference in an environment where the link can be interrupted or jammed.
Another major technical dimension is communications resilience: in the event of jamming, the drone continues its mission using its onboard logic, without relying on a human pilot or an active ground link. This enhances robustness in contested environments.
Civilian and military uses of real-time decision-making
Civilian uses
In precision agriculture, the drone identifies weeds and areas of hydraulic or sanitary stress in crops while in flight, then adjusts its trajectory to fly over or spray selectively. Onboard intelligence allows it to do this without returning to a ground station and optimizes data processing on site.
In infrastructure inspection (bridges, pipelines, wind turbines, high-voltage lines), the drone flies autonomously, automatically detects defects or anomalies (e.g., cracks, corrosion, intrusion), and adjusts its route to zoom in on the critical area. For example, the Portuguese company Tekever has developed surveillance UAVs with AI software for pipelines, migrants, and forest fires.
In logistics and delivery, autonomous drones equipped with Edge AI can identify obstacles, recalculate trajectories, drop off packages, and then automatically return to base. Onboard algorithms detect wires, trees, birds, and weather conditions. These capabilities enhance operational autonomy.
Military uses
In a military context, drones with onboard AI operate on several levels:
- ISR (Intelligence, Surveillance, Reconnaissance): the drone locates a target, tracks it, and shares essential data while continuing its mission.
- Autonomous or semi-autonomous engagement: the drone can identify an object of interest (vehicle, personnel, site) and propose an action or trigger a sensor without a constant connection. For example, Helsing AI offers an HX-2 attack drone capable of engaging armed targets up to 100 km away with onboard AI.
- Swarm and multi-UAV coordination: autonomous fleets carry out missions involving cover, bypassing, saturation of enemy territory, and large-scale threat detection. These locally operating fleets minimize communications and maximize efficiency. The article on MARL for UAVs demonstrated the effect.
- Operations in contested environments: electronic jamming, no GPS, severed connection. The drone continues its mission thanks to Edge AI, which is essential in future widespread conflicts. As noted in the Syntiant analysis, Edge AI enables action without delay and without dependence on communications.
- Autonomous real-time decision-making: detect an enemy, decide to avoid or engage, alert and act—all in a fraction of a second. The Markets & Markets report highlights that AI-drones are revolutionizing tactical and strategic capabilities.
Key players and financiers of this technology
Innovative companies
- Shield AI (USA): specializes in tactical autonomy for drones and aircraft via its Hivemind platform. It has raised hundreds of millions of dollars and is valued at approximately $5 billion in early 2025.
- Skydio (USA): autonomous drone for civil and military use, with inspection, security, and ISR missions.
- Edge Autonomy (USA/International): develops long-endurance UAVs with autonomous capabilities for ISR and defense.
- Palladyne AI (USA): software platform for autonomous, multi-agent, multi-sensor tactical drones.
- AgEagle Aerial Systems (USA): specializes in agriculture, but also autonomy and AI for drones.
- Tekever (Portugal): surveillance UAVs with AI for civil and military applications.
- Anduril Industries (USA): autonomous weapons and defense company, investing in AI for autonomous missions.
Financiers and financing issues
Financing is very important: billions are being invested in AI and drone autonomy. For example, according to one article, between 2019 and 2022, US military agencies allocated more than $53 billion to technology companies working on autonomous systems and AI. Shield AI’s fundraising also demonstrates the interest of private capital in this niche. The United States, China, Israel, and the European Union are increasing their research and development budgets for embedded AI, UAV autonomy, and robotic warfare. This is driven by the belief that victory goes to those who can think and act the fastest. (Edge AI)
The future of security and warfare with autonomous drones
The evolution towards autonomous drones with embedded intelligence is profoundly changing the nature of warfare and security. Several areas are worth highlighting.
Acceleration of the decision-making loop
Future conflicts will be marked by autonomous systems capable of detecting, deciding, and acting in a very short period of time. Autonomous drones no longer depend on a human pilot for every move. This shifts the advantage to those who can deploy fast, adaptive, and resilient fleets. In an environment where communications are jammed, the ability to operate in “link denial” becomes crucial.
Multiplication of fleets and swarms
Autonomous drones will operate in distributed swarms, coordinated by embedded algorithms. Each drone will make decisions locally while following an overall objective. These swarms can saturate enemy defenses, conduct deep reconnaissance, neutralize radars, and pave the way for heavier waves. The MARL study illustrates that this coordination reduces latency and energy consumption.
Increasing complexity of theaters of operation
With Edge AI, drones will be able to operate in urban, mountainous, and wooded environments, without GPS or with high levels of jamming. They will be able to classify targets, avoid protected areas, and replan in flight. Drones are becoming autonomous actors in complex theaters. The Markets & Markets report highlights that these capabilities extend overall operational efficiency.
Proliferation and defense challenges
As autonomous platforms become more commonplace, even less powerful actors can access autonomous drones. The asymmetric threat is increasing: non-state groups could deploy swarms of autonomous drones to harass or strike infrastructure. Defense will have to adapt by developing automatic countermeasures, detection sensors, and electronic warfare systems.
Questions of ethics, sovereignty, and human control
Autonomy implies decisions made by machines. This raises a fundamental question about the role of humans in the decision-making loop. If a drone identifies a “target” and acts on its own, who is responsible? States will need to define legal frameworks, “human-in-the-loop” architectures, and reliability guarantees. Furthermore, technological sovereignty is becoming central: who controls the algorithms, training data, and software chains? Major public funders are therefore vigilant about the origin of autonomous systems.
Plausible future scenarios
In the near future (2025-2030), we can envisage:
- autonomous drone patrols in border areas, detecting, classifying, and responding to threats without continuous human control;
- precision strike missions carried out by autonomous drones such as the HX-2 or equivalent, capable of operating beyond line of sight, with minimal connectivity;
- Civil-military fleets that closely share Edge AI technologies, making dual-use applications very common;
- Increasingly automated robotic warfare, where defense will have to rely on AI to detect and neutralize enemy autonomous drones.
Challenges to be overcome
Despite these advances, several technical and strategic challenges remain:
- Onboard computing power: having high-performance AI in flight, with low power consumption and limited weight, remains a constraint. NTT’s system is a good example of progress, but the challenge remains global.
- Data quality and robust algorithms: networks must learn to classify in varied conditions, with little supervision, and withstand the environment (weather, camouflage, jamming).
- Security, resilience, and certification: in a military environment, systems must be resistant to software or hardware attacks, interoperable, and reliable. The inclusion of an Edge Autonomy drone on the Defense Innovation Unit (DoD USA) “Blue UAS List” illustrates the certification requirements.
- Interoperability and tactical doctrine: Integrating autonomous drones into an existing chain of command and ensuring that they work with other platforms (manned aircraft, ships, satellites) requires doctrinal and organizational work.
- Regulation and ethics: The international community has not yet defined a common framework for the use of armed autonomous drones. This creates a risk of proliferation and misuse.
A strategic perspective
Edge AI technology for drone autonomy represents a paradigm shift similar to that of jet propulsion or stealth radar. Those who master decision speed, autonomous platform coordination, and real-time adaptation will enjoy a strategic advantage. In future warfare, the lines between surveillance, strike, logistics, and cyber operations are blurring: an autonomous drone equipped with AI can become a sensor, an attack vector, or a mission switch.
For armed forces and civil security actors alike, it will be a matter of anticipating, investing in Edge AI, building doctrines of use, ensuring autonomous logistics (maintenance, recharging, replacement), and developing countermeasures. Without preparation, a more agile actor could impose itself locally or even globally.
The issue is not only technological: it is political and moral. Who will decide when a drone acts? What assurance is there that autonomy will not lead to a loss of control? Proliferation risks taking these systems beyond the traditional sphere of the state.
In this landscape, every autonomous flight, every automatic classification, every unconnected trajectory adjustment becomes a strategic message. Drones with embedded intelligence are redefining not only how we fly, but who decides in the skies.
War Wings Daily is an independant magazine.