Inspiration

We were inspired by a simple question: What happens when robots lose communication with each other?

Most automated warehouses rely on centralised coordination to assign tasks and manage robots. But if communication fails, the entire operation can be disrupted.

For the Neptune North Challenge at GreatUniHack 2026, we wanted to explore a different approach: a warehouse where robots can think independently, cooperate with one another, and continue working even when parts of the network go down.

That idea became HiveMine.

What it does

HiveMine is an interactive, decentralised multi-agent warehouse simulation featuring five autonomous robots working together to fulfil customer orders.

Instead of relying on a central controller to assign every task, robots communicate with one another, evaluate available jobs, and negotiate task assignments.

Each robot makes decisions based on factors such as distance, battery level, workload and task priority.

What makes HiveMine different is its ability to demonstrate resilience during communication failures.

Users can:

  • Watch robots navigate the warehouse and complete orders.
  • Observe robots independently bidding for tasks.
  • Disconnect individual robots or split the network into separate groups.
  • Introduce communication delays and packet loss.
  • See how robots continue operating with their available local information.
  • Restore communication and observe how robots synchronise their knowledge.
  • Monitor performance, network activity and recovery through an interactive dashboard.

HiveMine demonstrates how a multi-agent system can remain functional even when communication is unreliable.

How we built it

We built HiveMine using Python, Streamlit and Plotly, combining multi-agent coordination algorithms with an interactive warehouse digital twin.

The system includes several main components:

  • Autonomous robot agents: Each robot maintains its own state, task knowledge, battery information and decision-making logic.
  • Decentralised task allocation: A CBAA-inspired bidding mechanism allows robots to negotiate work without a central task allocator.
  • Peer-to-peer network simulation: A simulated communication layer supports message exchange, network partitions, latency and packet loss.
  • Pathfinding and collision avoidance: A* pathfinding and conflict-based coordination help robots navigate the warehouse.
  • Battery and charging management: Robots monitor their energy levels and travel to charging docks when necessary.
  • Interactive dashboard: A visual interface displays robot movement, network connections, individual decisions and operational performance.

We also built automated tests to verify important behaviours, including task allocation, network failures, charging and recovery.

Challenges we ran into

One of our biggest challenges was making decentralisation genuine rather than simply making the system appear decentralised.

Robots needed to make decisions using their own information instead of relying on a shared controller that already knew everything.

We also encountered several technical problems.

Task allocation: Outdated task bids sometimes prevented robots from picking up abandoned jobs, causing the warehouse to stall.

Network failures: We had to ensure that disconnected robots genuinely stopped receiving messages from unreachable agents while continuing to make progress independently.

Battery management: Initially, robots could charge wherever they were. After implementing real journeys to charging docks, we discovered that some robots could run out of battery before reaching them.

Recovery: Restoring network connectivity did not automatically mean that agents had synchronised their information. We had to distinguish network reconnection from actual recovery.

Solving these issues helped us build a more realistic and reliable simulation.

Accomplishments that we're proud of

Our biggest accomplishment is turning decentralised coordination into something that people can actually see and interact with.

Rather than simply explaining distributed multi-agent systems, HiveMine allows users to experiment with them.

We're particularly proud of:

  • Implementing independent robot decision-making and decentralised task bidding.
  • Building a simulated network where communication failures have real consequences.
  • Creating an interactive warehouse digital twin.
  • Supporting task handoffs, charging dock allocation and network recovery.
  • Developing a guided demonstration that illustrates how the system responds to disruptions.
  • Improving reliability through repeated testing and debugging.

Most importantly, we built a system that demonstrates why decentralisation matters beyond theory.

What we learned

This project taught us that decentralisation is much more than removing a central controller.

Independent agents need reliable communication protocols, local decision-making, conflict resolution and recovery mechanisms.

We learned how difficult it is to coordinate multiple autonomous agents when they do not all have the same information.

We also gained practical experience with distributed algorithms, graph-based networking, simulation design, pathfinding, debugging and automated testing.

One of our biggest lessons was that restoring communication is only the beginning of recovery. Agents must also resolve differences between their local knowledge.

What's next for HiveMine

We see HiveMine as a foundation for exploring more advanced decentralised systems.

Our next steps include:

  • Improving the distributed task allocation algorithm.
  • Making robot navigation and collision avoidance more sophisticated.
  • Testing larger fleets with more complex communication networks.
  • Exploring adaptive coordination strategies for unreliable networks.
  • Introducing more realistic warehouse layouts and operational constraints.
  • Investigating integration with physical robots and real communication technologies.

Ultimately, we want HiveMine to demonstrate a broader idea: autonomous systems should not depend on perfect communication to keep working together.

Built With

Share this project:

Updates

Submission history