Anduril’s Lattice connects AI, mesh networking, and SDK to deploy interoperable autonomous capabilities as close to the field as possible, from sensor to effector.
In summary
Lattice is Anduril’s open operating system designed to bring autonomy and command & control to the tactical level. The architecture is based on a decentralized mesh network, Lattice Mesh, which securely distributes data and tasks between sensors, effectors, and command posts, even over long distances and in contested environments. For developers, Lattice SDK provides documentation, APIs, libraries (Go, Java, TypeScript, Python), examples, and sandbox environments to create decentralized applications, data services, and hardware integrations that can be pushed to the operational edge. Lattice is already being used to aggregate thousands of events, automate swarms, and orchestrate heterogeneous systems (air, land, sea) in real-world exercises and deployments. The U.S. Department of Defense has contracted for an industrial deployment of an “Edge Data Mesh” based on Lattice Mesh and has opened this mesh to third-party developers. The result is an open architecture platform that reduces decision-making latency, improves resilience to link failures, and accelerates the “sensor-decision-maker-effector” cycle.
The Lattice technological framework
Lattice is presented as an “operating system” for French-speaking defense tech: a single software layer that ingests streams from thousands of sensors, interprets them via AI, and extracts a common operating picture that can be used in a single interface. The goal is to compress the time between detection, classification, decision, and action. The platform is not tied to a specific vector: it controls or coordinates aerial systems (UAS, loitering munitions), naval systems (AUVs), land systems (sensor/effect turrets), and electronic warfare equipment. Lattice was introduced by Anduril as the basis for “Mission Autonomy”: a human can use it to plan, simulate, and supervise teams of autonomous assets across multiple domains.
Architecturally, Lattice isolates three blocks: multi-source ingestion, sensemaking (detection, tracking, correlation, intent estimation), and orchestration (task planning, order routing, security control). This separation allows components to be exchanged or added without disrupting the whole: the AI engine can evolve, as can the user interface, while maintaining compatibility with existing mission systems.
Distributed operation: mesh network, data, and tasks
The Lattice Mesh network
Lattice Mesh is a decentralized mesh network that transports data, events, and commands between services, platforms, domains, and long distances. It aims for resilience: no single point of failure, degraded operation during outages or jamming, automatic recovery when connectivity returns. The mesh supports dynamic topologies: mobile nodes (drones, vehicles), temporary relays, or satellite segments. The core promise: maintaining “useful consistency” of situation tables and task queues even in contested contexts.
The Department of Defense, through the Chief Digital and AI Office (CDAO), has entered into a three-year production agreement to “scale up” an initial “Edge Data Mesh” capability powered by Lattice Mesh; the initiative was then opened up to third-party developers to create applications that can be run on this mesh. This institutionalizes the idea of a tactical mesh network shared between agencies and industry.
The data model and task logic
Lattice exposes an entity model (assets, tracks, zones, threats, effects) and a task system: mission assignment, constraints, rules of engagement, navigation parameters, time windows. The API allows entities (e.g., a surface vessel) to be published, their status (telemetry, health, ammunition) to be tracked, and actions (monitor, track, identify, intercept) to be triggered. The platform distributes these tasks across available nodes, then aggregates the feedback into an event database. For the operator, abstraction protects against heterogeneity: radars, IR cameras, RF, LIDAR, all converge on common objects and behaviors.
Edge processing and resilience
Lattice implements edge computing logic: computing as close as possible to the sensor in order to reduce latency and lighten the load on connections. Local microservices perform filtering, detection, fusion, and prioritization; only useful artifacts are reported. In the event of an outage, the nodes retain models and rules to operate “offline,” then resynchronize. This approach enables swarm control, multi-sensor detection, and effect distribution even when throughput drops. Public demonstrations have validated these principles during joint exercises (e.g., coordinated control of autonomous assets during EDGE23).
The Lattice SDK: development, integration, deployment
The developer kit: documentation, API, libraries
Lattice SDK provides: documentation, getting started guides, API references, application examples, and sandbox environments. The canonical SDKs cover Go, Java, TypeScript, and Python, with official REST clients (entity publishing, task management, object/binary broadcasting). Installation consists of configuring authorization tokens, an environment endpoint, and then integrating the libraries into the project. The examples show how to publish an entity, attach sensors, and subscribe a service to events to react in near real time.
The typical development cycle
A standard cycle follows four steps:
- Describe entities and sensor/effect interfaces;
- Integrate the API to publish states and events;
- Compose tasks and rules (priorities, windows, thresholds);
- Deploy to the edge via the Lattice Mesh.
The Sandbox speeds up validation: pre-configured environments allow you to test the publication of a surface vessel or drone, simulate tracks, and then observe the routing of a task to an autonomous node. Developers can iterate without connecting to a classified system, reducing integration times.
Hardware and software integration
On the hardware side, Lattice interfaces with a wide family of Anduril and non-Anduril sensors and platforms. On the computing side, the Menace range (ruggedized computing) is now the preferred hardware for partner Edge stacks, facilitating the native execution of Lattice applications in compact, ruggedized enclosures for the field. On the software side, the open architecture and public APIs allow third-party services (video analysis, EW, SIGINT) to be connected if they comply with the entity/task model and security contracts.
Uses and missions: from Counter-UAS to multi-domain ISR
Site protection and Counter-UAS
In infrastructure protection, Lattice aggregates optical/IR, RF, and radar sensors, classifies tracks (UAS, birds, helicopters, rockets), and offers effect options (jamming, interception, kinetic neutralization). Command and control suggests sensor-effector couplings according to rules: for example, alerting a team, pointing a multispectral camera, then engaging an interceptor if the track crosses a restricted area. Publications have described large-scale automation of sensor towers in real environments, proving their scalability.
ISR and mission autonomy
On the ISR side, Lattice coordinates aerial patrols (UAS), rotary wings, surface effectors, and AUVs. Mission Autonomy allows complex missions to be defined: routes, search areas, assembly points, exclusion zones. A single operator can supervise dozens of robots; AI manages collision avoidance, role allocation, and weather adaptation. Lattice has been showcased in exercises where multiple heterogeneous assets cooperated to detect, track, and characterize targets, with dynamic replanning in the event of a node failure.
Institutional adoption and openness to third parties
Two key decisions: a three-year production agreement with the CDAO to industrialize an Edge Data Mesh within the DoD, followed by the opening of the mesh to third-party developers without going through the publisher. This confirms the shift towards shared platforms and interoperability at the tactical level. Units can embed their own “apps” (detection algorithms, trajectory calculators, fusion engines) and deploy them on authorized nodes.
Operational deployment: methods, metrics, feedback
Step-by-step commissioning
A typical deployment begins with a site survey: radio coverage, tolerable latency, mobility profiles, available power (12–28 V DC), thermal and IP constraints. The topology is then chosen: backhaul (satcom, microwave), mobile relays, local access points. Each node carries the minimum services: entity publication, event cache, rules engine. Integrators measure the refresh time of tracks, the probability of association in fusion, and the scheduling latency of tasks; Targets of < 1 s between classification and action recommendation are typical in close defense, while in large-area ISR, the constraint is more on persistence and link economy. (Indicative parameters depending on the sensor and link.)
Measuring military effect
Lattice’s value can be seen in three operational metrics:
— Shortened “sensor-effect” loop time (less than minutes, sometimes seconds);
— Rate of recommended actions accepted by operators;
— Success rate of interceptions/neutralizations or detection-pursuit.
Published cases show the automation of “hundreds” of sensor towers, and joint demonstrations confirm the conduct of multi-domain swarms. At the DoD level, the move to production aims for repeatability: same APIs, same entity model, same application packages to be propagated on the tactical edge.
Complexity and limitations: integration, cybersecurity, certification
Integration debt
Connecting existing fleets involves writing adapters to heterogeneous formats: ASTERIX, STANAG, MISB, and in-house protobufs. The Lattice SDK reduces the effort on the application side, but ontology alignment remains a work in progress. Tests must cover: packet loss, jitter, network partitions, redundancy of leading nodes, and entity identity collisions.
Cybersecurity and secret management
The decentralized model requires strict hygiene of secrets (tokens, certificates), a “least privilege” policy, and reliable logging for audits. In a multinational context, domain segmentation and data labeling (releasability) must be integrated into the modeling; Lattice provides the plumbing, while policies are the responsibility of the operational authority.
Certification and user acceptance
Accreditation cycles (RMF, NATO assessments, equivalents) require evidence: image hardening, SBOM, SAST/DAST scans, regression testing, and AI model traceability. The human aspect is just as important: the interface must remain simple, AI recommendations must be explainable, and margins for manual recovery must be clearly provided for, including in the event of node attrition.
Ecosystem dynamics: partners, hardware, software
Edge computing hardware
Recent announcements position the Anduril family of computers as the preferred hardware for partner Edge stacks. Advantages: turnkey deployment, cooling and hardening already resolved, and optimization for running Lattice Mesh microservices and third-party applications. For a program, this reduces integration times (weeks rather than months) and standardizes support in the field.
Third-party developers and standards
Opening up the mesh to non-Anduril developers has a strategic impact: it avoids proprietary lock-in, promotes algorithm competition, and paves the way for coalition interoperability. Public APIs, the Sandbox, and multi-language SDKs form the “bridge” between the defense world and the civilian software world. The open architecture logic is in line with MOSA principles: replaceable components, stable interface contracts, and clear separation of responsibilities.
Key takeaways for an armament program
For program managers, Lattice proposes an incremental modernization path: first connect the sensors and consolidate the common operating picture, then automate simple decision chains (detection-classification-alert), and finally introduce robust autonomy routines (patrols, pursuits, coordinated neutralizations). The benefits quickly become apparent if the scope is defined and measured by operational metrics. The risks relate to the alignment of ontologies, inter-agency security policies, and change management on the operator side; but the existence of industrialized Lattice SDK and Lattice Mesh, supported by a state production agreement, reduces the risk of dependence on fragile ad hoc integrations.
War Wings Daily is an independant magazine.