A public demonstration of a distributed team continuing useful work through communications loss, intermittent connectivity, agent failure, and newly emergent priority tasks by preserving mission intent and reallocating work locally.
Compared with 84.6% for the centralized assignment baseline across 100 deterministic synthetic trials.
The left panel uses a centralized assignment baseline: agents can finish their current task during a communications blackout, but cannot receive new assignments until connectivity returns. The right panel preserves mission intent locally and allows the surviving agents to continue tasking and redistribute work while disconnected.
Each trial uses eight agents, 36 routine tasks, two priority tasks that appear during degraded communications, one agent failure, a communications disruption and recovery schedule, and randomized task geometry from a fixed seed range. The live animation above is a separate representative scenario with six agents and 18 tasks so each behavior remains visually understandable.
Fraction of synthetic tasks completed by the end of each trial.
Reduction in mean completion latency for priority work introduced during communications loss.
Reduction in cumulative idle-agent steps while useful tasks remain available.
Reduction in assignment-request and assignment-response messages handled centrally.
Task completion is the fraction of all synthetic tasks completed by trial end. Priority-task response is the number of simulation steps from priority-task appearance to completion. Idle time sums agent-steps spent without an assignment while unfinished feasible work remains. Centralized assignment messages count assignment-request and assignment-response exchanges handled by the central coordinator.
Agents can finish work already assigned to them, but new tasking depends on reconnecting to the central assignment point.
Agents retain enough shared mission context to keep working when communications disappear and synchronize again when links return.
Distribute task definitions, priorities, role constraints, and basic team state while connectivity is available.
Each agent maintains its own position, health, current assignment, nearby task state, and communication status.
When links disappear, surviving agents continue selecting feasible work from locally available mission context.
Reassign unfinished work after agent loss and allow higher-priority tasks to interrupt lower-priority activity.
When communication returns, merge progress, resolve duplicate claims, and restore a common team picture.
Keep heterogeneous robots productive through intermittent links, partial partitions, and temporary loss of a team member.
Preserve objective-level mission behavior when centralized control or high-bandwidth connectivity is unavailable.
Support exploration, inspection, and monitoring where terrain or infrastructure makes communication intermittent.
Coordinate mobile robots or machines across large sites without turning a network interruption into a full mission stop.
Reallocate search, sensing, or delivery tasks when infrastructure is damaged and team membership changes dynamically.
Adapt assignments among vehicles or assets as connectivity, availability, and priority demands change.
Technical direction, networked/distributed system architecture, data and coordination interfaces, degraded-communications experiment design, reproducibility, performance evaluation, and end-to-end integration.
Multi-agent coordination, distributed task allocation, dynamic replanning, autonomous decision-making, confidence-aware AI, robustness analysis, and collaborative-behavior evaluation.
The demonstration is not submitted to an external autonomy benchmark. Its design reflects well-established research problems: intermittent communication, heterogeneous robot teams, task allocation, distributed planning, and resilience to changing team state.
NeBula describes autonomous coordination and task allocation among heterogeneous robots and resilient mesh networking designed to accommodate intermittent communication links.
Public reference ↗Recent 2026 research studies decentralized task allocation under uncertain task completion, dynamic arrivals, incomplete information, and sparse communication.
Public reference ↗