ADS-01
A swarm control system that combines ground robots and air robots. An AI agent proposes a plan of action based on the state of the environment. A human operator approves it.
Built for challenge 01 at the European Defense Tech Hackathon (EDTH) in Berlin, 27–29 November 2026.
The problem
Challenge 01 asks for one operator to task many different uncrewed vehicles at once. When one vehicle drops out, the plan has to change.
- One operator, many robots. Today, ground robots are often driven one at a time, one person per robot.
- Mixed fleet. Ground and air robots can do different jobs, have different batteries and lose their radio link at different times.
- Things go wrong. A robot loses its link or runs low on power, and its tasks need a new owner.
ADS-01 focuses on ground and air robots working as one team.
How it works
Ground robots + air robots→
State of the environment→
AI agent→
Proposed plan of action→
Operator approves, edits or rejects
- One map, one team. All robots share one map. Ground and air robots are run as one swarm, not as separate fleets.
- Tasks go to the best robot. The system scores each robot on what it can do, how far away it is, its battery and the quality of its radio link.
- The agent proposes. When something changes, the system proposes a new plan. It shows it as a list of changes, with the reason for each one.
- The operator decides. The operator approves, edits or rejects every change.
- Each robot keeps its own safety. A robot's own safety layer stays in charge of its wheels, whatever the plan says.
When the link drops
Radio links fail. Human control should not fail with them.
- Rules set in advance. Before the mission, the operator approves a fallback rule for each robot. For example: “if the link is lost for more than 30 seconds, hold, then return along the shared map.”
- Keep to the last approved plan. A robot that loses its link carries on with the last plan the operator approved.
- Check on reconnect. When the link comes back, the console compares what the robot did with what was approved.
The demo
The plan for the 48-hour hackathon:
- One real Movis ground robot, plus 6 to 10 simulated ground robots and 2 simulated drones, all on one map.
- Live, we cut the real robot's link or fake a low battery.
- The system proposes a new plan, with the reasons attached. The operator approves, edits or rejects it.
- Movis's own on-board decisions, each with its reasoning, show up in the same queue.
The goal: one operator runs a squad of ground robots, and every change to the plan is one they approved, with the reason attached.
Status
Research and design stage. The research behind this design was done on 22 November 2026. The build is planned for the hackathon, 27–29 November 2026. This page describes the planned design, not a finished system.
Risk: as of 21 November, the real robot’s drive base was not yet driving.