IN Brief:
- SwarmOS coordinated Group 1 and Group 2 aircraft supplied by multiple manufacturers.
- The software integrated with Anduril Lattice and the Army’s Next Generation Command and Control architecture.
- The exercise demonstrated a sensor-to-effects workflow, although no production award or detailed performance metrics were disclosed.
Palladyne AI has demonstrated multi-vendor drone control during the US Army’s Project Convergence Capstone 6 experiment at Fort Irwin, California.
Palladyne AI used its SwarmOS collaborative-autonomy software to coordinate heterogeneous Group 1 and Group 2 uncrewed aircraft from several manufacturers. The system integrated with Anduril’s Lattice platform, which served as an integration layer within the Army’s Next Generation Command and Control architecture.
The company also flew its Gremlin-X reusable uncrewed aircraft during the event. Palladyne said SwarmOS coordinated a sensor-to-effects sequence in which reconnaissance aircraft searched for targets, shared data through the Army command environment, passed a task to Gremlin-X after a soldier authorised engagement, and supported post-strike assessment.
The demonstration followed the 4th Infantry Division’s Ivy Mass rehearsal less than three months earlier. Palladyne said the later event increased platform diversity and mission complexity while placing the software inside a division-scale exercise involving US joint forces and allied participants.
Doug Dynes, president of Palladyne Aerospace & Defense, said: “SwarmOS delivered in both environments, enabling aircraft from multiple manufacturers to find targets, share information, adapt to battlefield changes, and coordinate precision effects while a soldier retained authority over the decision to engage.”
The distinction between controlling one autonomous aircraft and coordinating a mixed group is substantial. Individual autopilots can follow routes, stabilise flight, and execute local behaviours, but collaborative autonomy must allocate tasks across vehicles with different payloads, endurance, speeds, communication links, and levels of onboard processing.
Hardware independence is therefore the central claim being tested. A software layer that works with one manufacturer’s aircraft can be optimised around known interfaces; a multi-vendor controller must translate different data structures and command sets into a common operational model without reducing every platform to the least capable option.
That requires disciplined interface management. Position, health, fuel or battery state, sensor observations, mission status, and available actions must be represented consistently enough for the autonomy software to make useful decisions. When communications degrade, each aircraft also needs rules governing what it may continue doing, when it should wait, and how it rejoins the group after a link returns.
Palladyne said SwarmOS achieved Impact Level 5 certification for the exercise, covering higher-sensitivity unclassified data and mission-critical command-and-control workflows. Certification is not proof that every connected aircraft or mission application is secure, but it establishes a cloud-security baseline for the environment in which data and services are handled.
Project Convergence Capstone 6 gave the demonstration a broader systems context. The Army described the event as its largest Project Convergence experiment to date, involving about 10,000 participants and more than 90 technologies, concepts, and formations. Its central objective was continued validation of Next Generation Command and Control at division scale rather than a stand-alone drone competition.
That context matters because drone autonomy creates limited operational value when it cannot exchange data with the command architecture. A sensor track needs identity, confidence, location, timing, and authority information before another system can act on it. Passing those fields through Lattice and the Army environment while preserving a soldier’s decision role is more relevant to procurement than an isolated formation flight.
The demonstration also addressed operator burden. At Ivy Mass, Palladyne said one soldier supervised aircraft from multiple manufacturers after less than 30 minutes of training. The company has not disclosed how many aircraft were controlled at PC-C6, how often operators intervened, which communications were denied, or how performance was scored, so the scale of the result cannot yet be assessed independently.
Those missing metrics will matter in later evaluations. Useful swarm software should reduce the number of people required per aircraft without hiding faults or overwhelming the remaining operator with alerts. It should also degrade predictably when one vehicle, sensor, or network path fails rather than allowing a local problem to disrupt the entire mission group.
Multi-vendor autonomy presents a commercial question as well as a technical one. Open interfaces can let the Army add new aircraft and payloads without replacing the control layer, preserving competition between suppliers. Poorly defined interfaces can produce the opposite result, with each integration requiring bespoke engineering that recreates vendor lock-in through software services rather than airframes.
Palladyne has moved SwarmOS from rehearsal into a larger Army experiment, but the announcement does not amount to a production award. The next evidence should include repeatable performance measures, integration time for additional platforms, communications-loss behaviour, cybersecurity testing, and the manpower required to plan and supervise missions.
PC-C6 showed that the software could participate in a connected sensor-to-effects chain while a human retained engagement authority. Fielding will depend on whether the same control layer remains stable as aircraft numbers, vendors, mission types, and network constraints increase beyond a carefully supported exercise.
