Abstract:Building a distributed spatial awareness within a swarm of locally sensing and communicating robots enables new swarm algorithms. We use local observations by robots of each other and Gaussian Belief Propagation message passing combined with continuous swarm movement to build a global and distributed swarm-centric frame of reference. With low bandwidth and computation requirements, this shared reference frame allows new swarm algorithms. We characterise the system in simulation and demonstrate two example algorithms.
What problem does this paper attempt to address?
The core problem that this paper attempts to solve is to construct a fully - distributed, localized, and low - cost shared spatial reference frame for robot swarms. Specifically, the author aims to enable each robot to understand its position relative to the entire swarm only by relying on local observations and message passing, thereby achieving global spatial awareness. This will provide a basis for developing new swarm algorithms, especially in tasks that require global information, such as shape formation, logistics, etc.
### Specific Problem Description
1. **Spatial Awareness Problem**: In a robot swarm, a single robot can usually only perform local sensing and communication, so it is difficult to obtain global spatial information. However, many tasks (such as shape formation, environmental monitoring, logistics, etc.) require robots to have the spatial awareness ability of the entire swarm.
2. **Distributed Algorithm Design**: Traditional methods of obtaining global information often rely on centralized processing or assume that all robots have global position knowledge, which does not conform to the paradigm of distributed swarms. Therefore, how to infer and construct global information through pure local interactions is a key challenge.
3. **Low Bandwidth and Computational Requirements**: In order to enable this system to be widely deployed in practical applications, it is necessary to ensure that it has low bandwidth and computational requirements to adapt to resource - limited hardware platforms.
### Solution
The author proposes a method based on Gaussian Belief Propagation (GBP), combined with continuous swarm movement, to construct a globally - distributed spatial reference frame through local observations and message passing. Specific steps include:
- **Local Observation**: Robots observe nearby other robots through sensors and record relative position information.
- **Message Passing**: Using the GBP algorithm, messages are passed between robots and gradually converge to a consistent shared reference frame.
- **Factor Graph Construction**: Each robot constructs a local factor graph, which contains variable nodes representing 2D poses and measurement factors connecting these variables.
- **Dynamic Update**: As robots move, the factor graph is continuously updated, old nodes are removed, and new nodes are added to maintain the amount of information within the window.
### Application Examples
1. **Shape Formation**: Through the shared reference frame, robots can adjust their behavior according to pre - defined shape functions and finally form the required shape.
2. **Logistics Application**: Robots can use the shared reference frame to more effectively manage and retrieve the location information of cargo carriers, improving logistics efficiency.
Through this method, the author shows how to achieve efficient global spatial awareness while maintaining distributed characteristics, thereby making it possible to develop new swarm algorithms.