Abstract
Active sensors are essential for distance measurement in mobile robotics, yet they suffer from significant interference in multi-robot environments. This paper presents a decentralized navigation method that utilizes local communication to coordinate sensing timings. By implementing a time-shared scheduling scheme, robots can acquire accurate distance data without crosstalk. This data is used to construct local occupancy grids, which are then matched against a pre-existing global map to facilitate precise autonomous navigation.
Introduction
In multi-robot systems, the ability to sense the environment accurately is a prerequisite for coordinated movement and task execution. Active sensors, such as ultrasonic and infrared sensors, are commonly used for distance estimation. However, when multiple robots operate in proximity, their signals often overlap, leading to "crosstalk." This interference results in erroneous distance readings and "ghost" obstacles, making reliable navigation nearly impossible. This paper addresses this challenge by treating the sensing window as a shared resource managed through local communication.
Methodology
Time-Shared Scheduling via Local Communication
To mitigate sensor interference, the authors propose a time-division multiplexing approach. Rather than relying on a centralized controller, robots communicate locally to negotiate sensing windows. When a robot needs to perform a distance measurement, it "occupies" the local temporal space. During this brief interval, neighboring robots refrain from firing their active sensors. This decentralized scheduling ensures that only one sensor is active in a specific vicinity at any given time, effectively eliminating the risk of catching reflections from other robots' signals.
Navigation through Map Matching
Once clean distance data is obtained through the scheduled sensing windows, the robot constructs a local occupancy grid. This grid represents the environment as a series of cells, each containing a probability value indicating whether the space is occupied.
To determine its position and orientation, the robot performs a matching process between its real-time local occupancy grid and a pre-loaded global map of the environment. By identifying the transformation that provides the best fit between the two maps, the robot can localize itself with high precision. This grid-based approach offers robustness against minor environmental changes, as the matching algorithm focuses on the most consistent static features of the global map.
Key Insights and Discussion
- Temporal Space Occupancy: A primary contribution of this work is the concept of treating time as a resource that can be "occupied" just like physical space. By managing time-shared access to the environment, the physical limitations of active sensors are bypassed.
- Decentralized Scalability: The use of local communication instead of a global network enhances the system's scalability. Robots only need to coordinate with immediate neighbors within their interference range, rather than every agent in the system.
- Robust Localization: Using occupancy grids for matching is more resilient than simple point-cloud comparisons. It allows the system to handle noise and small discrepancies between the real-time environment and the static global map.
Conclusion and Future Work
The proposed method successfully demonstrates that time-shared scheduling can resolve sensor crosstalk in multi-robot systems, enabling reliable navigation through map matching. However, certain limitations remain. In high-density environments, the waiting time for a sensing slot may increase, potentially slowing down robot movement. Additionally, the system currently relies on a static global map. Future research in this lineage explores multi-robot Simultaneous Localization and Mapping (SLAM), where robots can build and update shared maps in real-time within dynamic environments.
Summary updated