# Lockheed Martin Skunk Works Proves AI Can Replan Drone Missions in Seconds in Autonomous UAV Demo

> AI flagged a fuel issue mid-mission, offered replan options in seconds, and a human picked the best. It re-tasked another drone and kept the mission on track.

- Type: AI news
- Page: https://completeaitraining.com/news/lockheed-martin-skunk-works-proves-ai-can-replan-drone/
- Published: 2025-12-05
- Updated: 2025-12-05
- Author: Jeroen Erne
- Jobs: Management
- Reading time: 3 min

### AI That Solves Mission Problems in Real Time: What Managers Need to Know

Lockheed Martin Skunk Works® ran a live demo of AI-driven mission-contingency management on a Stalker XE Block 25 UAV and a modified Alta X 2.0 drone. The system detected a fuel issue, generated re-plan options in seconds, and let the operator approve the best one. Once selected, the AI re-tasked the Alta X to finish the job and sent the Stalker back to base. Human judgment stayed in the loop; the AI handled the heavy lift.

#### Why this matters for leadership

- Operational resilience: missions keep moving even when assets fail or conditions change.
- Speed: decision latency drops from minutes to seconds, especially during off-nominal events.
- Workforce leverage: operators supervise and approve, while AI handles re-planning and task hand-offs.
- Interoperability: one node coordinated air and ground assets, signaling lower integration overhead.
- Risk control: clear approval points preserve accountability and reduce error cascades.

#### How the demo worked

The exercise simulated multiple fuel contingencies on the Stalker XE. The ground C2 AI assessed the situation, produced re-plan options, and surfaced them to the operator via a simple interface. After the operator chose a path, the AI shifted tasks to the Alta X and commanded the Stalker to return to base. The outcome: faster recovery with the operator focused on mission priorities rather than manual re-tasking.

#### One command node, air and ground

The Stalker fed mission data to a unified C2 node that also managed an Unmanned Ground Vehicle in Kansas, with support from Fulcrum-provided UAVs. This showed cross-domain control from a single, mobile command node. It directed multiple, geographically separated drone meshes in mounted, dismounted, and below-the-noise setups, which points to scalable control across dispersed operations.

#### The tech behind the hand-off

The team used STAR.SDK™ within the STAR.OS™ constellation to plug the contingency app into a chat-based interface for operator approval and re-tasking. That same foundation enabled different AI systems to interoperate across unmanned platforms without bespoke wiring each time. More on multi-domain integration is available at Lockheed Martin's site: [Multi-Domain Operations](https://www.lockheedmartin.com/en-us/capabilities/multi-domain-operations.html?ref=completeaitraining.com).

#### Leadership playbook: first 90 days

- Run a tabletop on your top five contingencies (fuel, comms loss, sensor failure, weather, geo-fence breaches) and define who approves which automated actions.
- Set human-on-the-loop thresholds: which events can auto-execute with notification, which require a quick confirm, which mandate a hold.
- Pilot within a low-risk mission segment; collect data on re-plan speed, mission continuity, and operator workload.
- Demand open interfaces for C2 and telemetry so you can mix vendors without lock-in.
- Update SOPs: escalation paths, fallback modes, and training for supervisors to audit AI decisions.

#### Risks and guardrails

- Model reliability: validate on real-world edge cases, not just clean lab scenarios.
- Comms denial: ensure degraded-mode behaviors and delayed-sync logs for after-action review.
- Cybersecurity: treat the C2 node as a tier-one asset; require secure boot, signed updates, and continuous monitoring.
- Authority and ethics: retain human approval for target changes and mission aborts; align with published principles such as [DoD AI ethical principles](https://www.defense.gov/News/Releases/Release/Article/2091996/dod-adopts-ethical-principles-for-artificial-intelligence/?ref=completeaitraining.com).

#### Metrics that matter in pilots

- Time from contingency detection to approved re-plan.
- Mission completion rate after a contingency event.
- Operator interventions per flight hour and per asset.
- Operator workload (e.g., NASA-TLX or equivalent) before vs. after AI/MCM.
- Comms bandwidth used during re-plan and task hand-off.

#### Procurement checklist

- Evidence of cross-vendor UxV control from a single C2 node.
- APIs, data schemas, and SDK support (documentation, SLAs, version cadence).
- Built-in audit logs: what the AI proposed, what the operator approved, and timing.
- Simulation hooks for scenario stress-tests and red-team exercises.
- Fallback behaviors: hold, return-to-base, safe loiter, and human override.

#### What the team said

"This demonstration proves AI can move from the lab to the battlefield, delivering a multitude of capabilities ranging from autonomous decision-making to rapid data flow between unmanned vehicles across air, ground and synthetic environments," said OJ Sanchez, vice president and general manager, Lockheed Martin Skunk Works. "By fusing AI-enabled UAV replanning with UGV capabilities, we give warfighters the safety, speed and confidence they need to act first in contested environments."

#### Take the next step

If you're standing up internal capability, review curated training by role: [AI courses by job](https://completeaitraining.com/courses-by-job/). Equip supervisors and mission owners to read, challenge, and approve AI-generated plans with confidence.

---

This is the Markdown form of https://completeaitraining.com/news/lockheed-martin-skunk-works-proves-ai-can-replan-drone/. The whole site is mapped at https://completeaitraining.com/llms.txt.
