AI, GPS-free navigation, and collaborative combat: state of the art, key manufacturers, open architectures, and capability challenges for the sixth generation.
Summary
Sixth-generation fighter aircraft and combat drone programs are relying on onboard artificial intelligence to increase range, survivability, and military effectiveness. AI optimizes multi-sensor data fusion, guides GPS-free navigation in jammed environments, and pilots collaborative combat drones controlled by a lead fighter (manned-unmanned teaming). Concrete milestones are already in place: close combat flights by an AI-controlled F-16 (DARPA ACE), selection of Anduril and General Atomics for the first CCAs, and the rise of open architectures (MOSA/OMS, ISANKE & ICS). Airbus is preparing Wingman, MBDA its Remote Carriers, while HENSOLDT, Thales, and Leonardo are industrializing cognitive electronic warfare. Expected benefits: millisecond decision times, massification of effectors, reduced risks for the crew. But AI requires firm choices on cybersecurity, rules of engagement, and software sovereignty.
The state of the art in embedded AI
The 6th generation is not just about a stealth airframe: the differentiator is software. Embedded artificial intelligence lightens the pilot’s cognitive load, optimizes tactical trajectories, prioritizes threats, and proposes firing solutions in near real time. In April 2024, DARPA’s ACE program conducted real-world dogfight flights with an AI-piloted F-16 (X-62A VISTA) against a human-piloted F-16, validating non-deterministic algorithms under safety supervision. The engagements reached a separation of ~610 m at approximately 1,930 km/h (1,200 mph), a world first at this level of demand, intended to build confidence in combat autonomy. These tests set out the roadmap: AI as a tactical co-pilot, with humans making critical decisions.
List of priority AI building blocks
– Decision support: automatic track classification, probabilistic assessment of effects, sensor/weapon priority management.
– Mission execution: synthesis of a single tactical image, anomaly detection, dynamic replanning.
– Tactical control: optimized flight envelopes, consumption/range trade-offs, evasive maneuvers.
– Human-machine collaboration: delegation of tasks to drones, sharing of intentions and constraints.
– Cybersecurity: model hardening (adversarial ML), integrity control, fault tolerance.
GPS-free navigation and resilient PNT
Contested theaters render GNSS degraded or denied. GPS-free navigation combines inertial navigation systems, radar altimetry, vision/terrain assistance, stable clocks, and anti-jamming M-Code receivers. Honeywell refers to layered “resilient navigation”: jamming detection, anti-spoofing, alternative sensors, and INS/GNSS integration. BAE Systems is expanding its NavStorm+ range (M-Code, anti-jam, integration with INS) for ammunition, drones, and aircraft. The USAF has approved a modular MOSA R-EGI (Resilient-EGI) architecture capable of integrating alternative third-party PNT. The goal is to maintain a metric CEP despite jamming. This resilience is essential for low-altitude flight, refueling, and coordinated stand-off strikes.
Specific technical choices
– Ruggedized G multi-channel M-Code receivers (missiles, drones, pods).
– Multi-hypothesis filtering and visual odometry to maintain navigation under jamming.
– Resonator/optical clocks and quantum sensor testing (DoD program) to improve time keeping.
Data fusion and open architecture
Data fusion requires high-density computers and low-latency buses, under MOSA/OMS, in order to quickly integrate new sensors/algorithms. On the European side, GCAP has industrialized ISANKE & ICS — ISANKE & ICS — which combines active/passive sensors, links, and non-kinetic effects within a single electronic stack; integration flights are planned on the Excalibur test aircraft. The approach aims for a distributed tactical cloud with low probability of interception, with embedded services for electronic warfare, detection, and C2. In the United States, OMS/JADC2 connects platforms and effectors via open interfaces tested in joint exercises. The challenge is as much technical as it is contractual: reducing vendor lock-in and integration cycles from several years to a few quarters.
Operational implications
– Longer-range detection through multi-band correlation (AESA, IRST, ESM).
– Real-time sharing of reliable tracks and tactical intent.
– Rapid software reconfiguration (new threats, EW libraries).
Collaborative combat and CCAs
At the heart of the 6th generation architecture is manned-unmanned teaming between a “leader” aircraft and collaborative combat drones (Collaborative Combat Aircraft). In 2024, the USAF selected Anduril and General Atomics to manufacture and test “Increment 1” CCAs, with a production decision targeted for 2026. The USAF plans to acquire more than 100 CCAs over five years for the initial phase, and aims to eventually have a fleet of around 1,000 units; The target unit cost is estimated at around $30 million (~€28 million) according to open estimates. Recent prototypes include the XQ-67A (designated YFQ-42A) and the Anduril platform (YFQ-44A). Typical missions: sensor scout, EW escort, weapons carrier, active decoys, penetrating strike.
Players and capabilities
– General Atomics: XQ-67A, “genus/species” concept to industrialize variants on a common chassis.
– Anduril: Lattice (mission autonomy) and collaborative aircraft, selected for Increment 1.
– Boeing: MQ-28 Ghost Bat in Australia; USN/USAF activities.
– Kratos: XQ-58A Valkyrie in limited production, partnership with Airbus for German capability by 2029; announced range ~4,800 km (3,000 miles), ceiling ~13,700 m (45,000 ft), ramp launch.
– Europe: Airbus Wingman (AI co-development with Helsing), MBDA Remote Carriers (ERC ~4 m, ~400 kg). These building blocks add mass and dilute risk, at one-third the cost of a modern fighter jet according to Airbus.
Cognitive electronic warfare and self-protection
Cognitive electronic warfare relies on AI to recognize new waveforms, adapt jamming, and update libraries during missions. HENSOLDT is advancing its Kalaetron family (ESM/EA) with signal recognition and multiple attack modes. Thales is modernizing SPECTRA (Rafale) and preparing AI image analysis capabilities in the TALIOS pod, reducing dependence on external links and accelerating designation. These functions are integrated into integrated sensor architectures (e.g., ISANKE & ICS on GCAP) to combine detection, self-protection, and non-kinetic effects in a single software continuum.
Examples with figures
– Broadband emission recognition and in-mission learning (partial library update).
– Multispectral detection and tracking: RF/IR/laser correlation, reduced detection time and false alarms.
– AI pods: edge processing for real-time identification without relying on network feedback.
The AI co-pilot and the mission interface
In the cockpit, AI becomes the co-pilot: it suggests maneuvers, manages drone swarms, and prioritizes information. The ACE tests sought to measure the crew’s confidence in machine decisions, from defense to nose-to-nose attack. This standardization of human-AI interaction is a prerequisite for large-scale deployment: autonomy modes, disengagement criteria, logging, and post-mission review. Players such as EpiSci (Tactical AI), Shield AI (embedded artificial intelligence “Hivemind”) and Anduril (Lattice) are validating components in flight on light jets and fast drones.
Predictive maintenance and software security
AI is extending to support: early fault diagnosis, intelligent maintenance planning, and spare parts inventory optimization. The savings are measured in operational availability, but require robust telemetry, secure data pipelines, and audited models. On the security side, embedded systems must be able to withstand attacks on models (adversarial), data poisoning, and GNSS jamming. Authorities will require formal assessments (model traceability, robustness testing, minimum explainability for critical functions).
Strategic consequences
The first consequence is the mass production of effectors with lower unit costs. CCAs costing around $30 million (~€28 million) make it possible to accept risks that would not be acceptable with a fighter jet costing hundreds of millions. The second effect is decision speed. A distributed combat cloud network, with edge computing, compresses the detect-designate-shoot loop and reconfigures the mission in flight. Third impact: software dependency. Forces that control their AI stack (data, training, software MCO) will avoid dependency on single providers. Finally, rules of engagement will have to regulate delegated use: humans “on-the-loop” at a minimum, tamper-proof logs, and safe shutdown procedures in case of doubt.
The market, manufacturers, and technological positioning
United States: USAF NGAD (family of systems) and USN F/A-XX are betting on a lead fighter and a CCA constellation. Anduril and General Atomics are leading Increment 1, while Boeing, Lockheed Martin, and Northrop Grumman retain a role in other areas (adaptive engines, sensors, OMS/JADC2). Europe: GCAP (United Kingdom, Italy, Japan) is industrializing ISANKE & ICS; Airbus with Helsing is pushing Wingman, MBDA its Remote Carriers; HENSOLDT, Leonardo, and Thales are densifying electronics and EW. Germany is exploring a rapid CCA capability with Kratos XQ-58A integrating an Airbus mission system, targeting 2029. These trajectories embody a global convergence: open architectures, distributed AI, and swarms piloted by a leader.
The 2025-2040 capability trajectory
2024-2026: AI maturation (ACE), CCA demonstrators (YFQ-42A/YFQ-44A), first ISANKE/ICS techno flights, R-EGI and alternative PNT trials. 2027-2030: CCA pre-production, integration of manned-unmanned teaming into transition squadrons, AI pods in operations, deployment of the first collaborative combat drones in coalition. 2030-2035: doctrinal shifts (delegation of penetration tasks to drones), maturation of cognitive electronic warfare, improvement of multi-spectral survivability. 2035-2040: entry into service of 6th generation families with interoperable combat cloud, swarms of modular-cost effectors, and a growing share of technical decisions delegated under human supervision.
A red line and measured bets
AI does not dispense with clear choices. Forces that want to exploit 6th generation fighter aircraft will have to finance training data, standardize MOSA/OMS across the entire fleet, and accept constant cybersecurity efforts. Countries that “buy” AI as a block will freeze their capabilities. Those that co-design it, with capable partners (Anduril, General Atomics, Airbus/Helsing, MBDA, HENSOLDT, Thales, Leonardo, Northrop Grumman) will maintain their advantage. The rest boils down to two straightforward technical questions: where does autonomy reside—on board the vehicle or in the combat cloud—and who controls the software supply chain, from model to deployment?
Live a unique fighter jet experience