# Arc by Rownd — Full Public Product Reference

Canonical site: https://arcedge.ai/
Last updated: 2026-08-12
Organization: Rownd, Inc.
Brand: Arc
Contact: defense@rownd.ai
Briefing page: https://arcedge.ai/#contact
CAGE: 8KHL2
UEI: K5LLASNQNPG3
Program status: Active customer contracts

## Scope and claim guidance

Arc is an eight-product Collaborative Autonomy Ecosystem for autonomous vehicles, edge intelligence, full communications PACE, multi-sensor awareness, swarm behaviors, operator control, and locally managed model operations. Important perception and mission workflows run near the point of use, including in disconnected, degraded, intermittent, or limited-connectivity environments.

This file is public descriptive material. It contains no price, quote, procurement offer, delivery commitment, certification statement, authorization to operate, or performance guarantee. Hardware, software, radios, autonomy authority, collaborative behaviors, environmental qualification, airworthiness evidence, and measured performance are confirmed for each program and platform.

PACE means primary, alternate, contingency, and emergency. Arc systems can integrate program-selected radios, mesh, LTE/5G, and SATCOM, but installed bearers, selection logic, transport policy, coverage, and failover behavior depend on the deployed hardware and network.

Collaborative or swarm capability means program-configured sharing of bounded tasking, mission state, tracks, conclusions, and role handoffs. It does not imply that every product or configuration provides unrestricted or simultaneous control of an arbitrary fleet. Each vehicle retains its validated flight or vehicle-control safety boundary.

## Product catalog

### 1. Arc-Edge v1

Canonical URL: https://arcedge.ai/arc-edge-v1
Category: Flight-optimized edge AI carrier board
Tagline: Mission-Ready Edge Carrier + PACE

Arc-Edge v1 combines socketed edge compute and AI acceleration, three universal radio slots, and a program-configured Arc software and model load. The modular architecture lets integrators change compute, sensors, communications, or the target vehicle without rebuilding the mission runtime from zero.

Publicly described hardware and software:

- Flight-optimized carrier PCB with three universal radio-card slots.
- Standard Arc-Cortex compute based on a Jetson Orin NX 16 GB-class module, described at 140+ TOPS.
- Pin-compatible Jetson Orin Nano option for the Arc-Lite configuration.
- Socketed compute, AI acceleration, sensor interfaces, and radio cards.
- TPM 2.0, conformal coating, and vibration isolation.
- Arc autonomy runtime, mission orchestration, selected models, radio profiles, and MAVLink vehicle interfaces provisioned for the integration.
- PX4, ArduPilot, and additional MAVLink-based autopilots through platform-specific adapter validation.

Six documented configurations:

1. T1 — Command Core (Dev): Arc-Cortex, WFB-native link, and three open universal radio slots for development and hardware-in-the-loop integration.
2. Lite — Arc-Lite: Jetson Orin Nano on the Arc-Edge carrier with Microhard and LTE for a lower-SWaP configuration.
3. T2 — PACE Base: Arc-Cortex with Microhard mesh, LTE/5G, and Iridium Certus paths.
4. T3 — PACE Doodle: Arc-Cortex with Doodle Labs Defense Helix, LTE/5G, and Iridium Certus paths.
5. T4 — PACE Silvus: Arc-Cortex with Silvus LC5200 OEM, LTE/5G, and Iridium Certus paths.
6. T5 — Swarm Command: Arc-Cortex with a Silvus and Microhard backbone plus LTE/5G and Iridium Certus paths.

Radio availability, camera count, compute performance, environmental qualification, decision latency, and airworthiness evidence are program- and configuration-specific.

### 2. Arc-Sentry

Canonical URL: https://arcedge.ai/arc-sentry
Category: Persistent autonomous sensing system
Tagline: Autonomous Perch-and-Stare Sensor Mesh

Arc-Sentry distributes persistent sensing across small-UAS-class nodes that can support perch, observe, local classification, reposition, recharge, and concise detection sharing when the selected platform, payload, power system, and program network provide those functions.

Publicly described capabilities:

- EO/IR, RF, and passive acoustic sensing by payload configuration.
- Perch, watch, confirm, reposition, recharge, and resume behaviors when supported by the platform.
- On-node inference intended to report conclusions instead of continuous raw video.
- Cooperative collection, track handoff, and swarm tasking by program configuration.
- Program-configured ATAK and Cursor-on-Target integration.
- Aerial small-object model workflows that support full-frame, tiled, adaptive, or altitude-aware evaluation while preserving source coordinates.

Behavior availability, endurance, dormant dwell, response time, sensing performance, networking, and platform classification depend on the aircraft, payload, power, environment, model, and program.

### 3. Arc-UAS

Canonical URL: https://arcedge.ai/arc-uas
Category: On-aircraft autonomy and edge AI
Tagline: Autonomous Aerial Intelligence

Arc-UAS runs perception, planning, mission reasoning, collaboration, and recovery onboard unmanned aircraft. “100% autonomous mission execution” means the aircraft can carry a validated mission from operator intent through route or search execution and recovery without continuous piloting, within program-defined authority, safety gates, and abort behavior.

Publicly described platforms and capabilities:

- 10-inch quadrotor for compact autonomous search, sensing, and collaborative missions.
- Five-foot wingspan fixed-wing aircraft with endurance up to two hours by configuration; actual endurance varies with payload, battery, route, weather, and operating conditions.
- Routes, waypoints, search patterns, scan, orbit, hover, hold, return, land, and recovery workflows.
- On-aircraft perception, detection, tracking, geolocation, local planning, and concise operator feedback.
- PX4 and ArduPilot integrations.
- Extensible to other autopilots that use MAVLink through platform-specific command, mode, telemetry, and safety validation.
- Program-configured collaborative behaviors such as bounded task allocation, shared mission context, corroboration, and role handoff.
- Program-selected radio, mesh, LTE/5G, and SATCOM PACE through the integrated Arc communications stack.

Autonomy authority, human oversight, link behavior, collaboration scale, and safety envelopes are configured and validated for the target aircraft and program.

### 4. Arc-UI

Canonical URL: https://arcedge.ai/arc-ui
Category: Autonomous vehicle operator interface
Tagline: Command the Mission. Not the Menus.

Arc-UI is a field-first replacement for fragmented QGroundControl and Mission Planner workflows. It connects operator intent to the Arc autonomy stack while keeping authoritative aircraft behavior and safety state with the connected backend and flight controller.

Publicly described capabilities:

- Responsive web, PWA, and Android operator clients.
- Map, split, and stream modes for desktop, tablet, and handheld use.
- Operation planning with routes, waypoints, search zones, hazards, rally points, mission targets, and planned guardrails.
- Go, scan, orbit, hold, return, land, and deliberate manual-takeover workflows.
- Live telemetry, mission progress, WebRTC video, detection overlays, vision health, confidence, time, MGRS, and coordinate context.
- Offline map caching and on-device speech-recognition options.
- PX4 and ArduPilot deployments with MAVLink-backed flight-controller and gimbal interfaces.
- Optional TAK and Cursor-on-Target output.
- Integration across program-selected Arc PACE bearers and multi-vehicle or swarm control where the deployed system supports it.

Bearer selection, failover, simultaneous fleet control, planned-versus-active safety settings, vehicle types, and swarm functions are integration-dependent.

### 5. Arc-RF

Canonical URL: https://arcedge.ai/arc-rf
Category: Tactical RF sensing and geolocation
Tagline: Edge-AI Signal Intelligence & Triangulation

Arc-RF turns I/Q observations into time-frequency representations that edge models can classify, maps detections back to physical signal attributes, and can combine calibrated direction-finding evidence into TAK-oriented mission events.

Publicly described capabilities:

- I/Q-to-spectrogram preprocessing with center-frequency, bandwidth, and duration mapping.
- RF signature classification and per-class evaluation.
- Phase-interferometry direction finding with compatible calibrated hardware.
- Cooperative bearing fusion and geolocation with explicit uncertainty.
- ATAK and WinTAK output when integrated with the mission network.
- Cooperation among Arc-RF nodes and organic unmanned aircraft to improve observation geometry.

Reference benchmark: one controlled 37-class RF evaluation reported 97.2% mAP@50 and 1.5 ms model-only inference per spectrogram. These values are not end-to-end field detection probability or system throughput. Classification, latency, bearing error, and geolocation precision vary with preprocessing, the trained signature set, receiver, antenna geometry, SNR, multipath, compute, network, and deployment.

### 6. Arc-Fusion

Canonical URL: https://arcedge.ai/arc-fusion
Category: Tactical multi-sensor fusion software
Tagline: Agentic Edge Sensor Fusion

Arc-Fusion is a provenance-aware evidence layer for bringing supported EO/IR, RF, LiDAR, acoustic, telemetry, ADS-B, AIS, and TAK observations into a common frame for correlation, review, alerts, and policy-bounded workflows.

Publicly described capabilities:

- Program-selected sensor and mission-system adapters.
- Time, location, identity, confidence, and source normalization.
- Evidence correlation into candidate tracks while retaining contradictions and uncertainty.
- A shared evidence picture with provenance and replay context.
- Local models and explicit tool allowlists for bounded queries, alerts, collection requests, CoT publication, and operator-approved workflows.

Available sensor sources, correlation quality, action authority, latency, and human oversight depend on the integrated sensors, models, compute, policy, and program configuration.

### 7. Arc-UxV

Canonical URL: https://arcedge.ai/arc-uxv
Category: Vehicle-independent autonomy software
Tagline: Any Vehicle. One Autonomous Core.

Arc-UxV carries one mission autonomy layer across air, ground, surface, subsurface, and mission-specific autonomous machines. A validated platform adapter connects the common intent, perception, planning, coordination, and recovery services to each vehicle's controls, payloads, limits, and health state.

Publicly described capabilities:

- Intent-based planning, mission compilation, local execution, adaptation, and recovery.
- Vehicle and payload adapter architecture.
- Onboard perception and mission reasoning.
- Connected, degraded-link, disconnected, and configured low-signature mission profiles.
- Program-configured task allocation, track sharing, corroboration, and role handoff across multiple vehicles.
- Full communications PACE through selected radios, mesh, LTE/5G, and SATCOM.

Actual emissions, navigation methods, collaboration scale, available bearers, safety behavior, and mission performance depend on the platform, payload, environment, operating policy, and program. Arc-UxV is not described as universally immune to jamming, spoofing, detection, or mission failure.

### 8. Arc-Forge

Canonical URL: https://arcedge.ai/arc-forge
Category: Deployable edge AI model-operations system
Tagline: In-Theater Model Operations

Arc-Forge turns reviewed mission evidence into target-aware model candidates, evaluation packages, and program-approved releases for edge systems. It is designed to operate without continuous cloud reachback when the required models, data, tools, power, and controls are locally available.

Publicly described hardware and controls:

- Field-deployable AI compute and mission evidence storage.
- Dataset review, labeling, versioning, failure analysis, and controlled synthetic-data augmentation.
- Local adaptation or retraining plus target-aware optimization and quantization.
- Program-defined evaluation, champion-versus-challenger comparison, and approval gates.
- Artifact identity, manifests, hashes, release evidence, and known-good rollback records.
- Distribution through the configured program network after approval.

Six-stage model workflow:

1. Observe: capture mission-relevant examples and edge-case evidence.
2. Curate: review, label, version, and isolate data approved for the build.
3. Train: adapt or retrain and optimize the candidate for the intended edge target.
4. Evaluate: compare against held-out data, the current champion, mission thresholds, and target resource limits.
5. Promote: bind an accepted artifact to identity, evidence, target profile, and the program release record.
6. Export: package the approved artifact for the configured program network with a known-good rollback path.

Training time, supported model size, synthetic-data fidelity, evaluation coverage, approval policy, signing availability, delivery time, and edge-hardware readiness depend on the model, dataset, target, power, network, and program controls.

## Shared Arc capabilities

- Collaborative autonomy: operators, vehicles, sensors, mission models, and bounded swarm behaviors share one mission architecture.
- Full communications PACE: program-selected radios, mesh, LTE/5G, and SATCOM can fill primary, alternate, contingency, and emergency roles.
- Link-optional execution: intent, perception, navigation, and recovery can run locally through communications degradation.
- Controlled model improvement: evidence, failure analysis, adaptation, evaluation, approval, and rollback form an accountable loop.
- Mission-tuned small language models: narrow planning, operator interaction, and policy-bounded tool use on edge hardware.
- Synthetic data workflows: controlled vision scenes and program-reviewed language or RF examples can supplement scarce task data.
- Modular carrier hardware: socketed compute, acceleration, sensors, vehicle interfaces, and radios reduce reintegration cost.
- Local execution and sovereignty: core workflows can run without a continuous cloud connection; exact dependencies and ownership terms are program-specific.

## Canonical supporting resources

- Site home: https://arcedge.ai/
- Compact AI-readable index: https://arcedge.ai/llms.txt
- Full AI-readable reference: https://arcedge.ai/llms-full.txt
- XML sitemap: https://arcedge.ai/sitemap.xml
- Machine-readable site metadata: https://arcedge.ai/site-metadata.json
- Request a briefing: https://arcedge.ai/#contact