TL;DR
This paper introduces a novel middleware for enabling multiple robots to collaboratively accompany a human, extending the capabilities of existing single-robot systems. The middleware facilitates swarm-driving by managing robot coordination, communication, and relative positioning, allowing for more robust and versatile human-support services.
What Problem Does This Paper Address?
The increasing demand for human-supporting robots has led to the development of systems like "Pet-Bot," which can assist with walking and carrying objects. However, these systems are limited by the number of robots that can be controlled by a single user, restricting their application in scenarios requiring the transport of multiple or heavy items. This paper addresses this limitation by proposing a "swarm-driving" approach, where multiple robots cooperate to achieve a common goal – accompanying a human – by mimicking the efficient parallel processing and optimal method detection seen in natural swarms. The challenge lies in coordinating these robots in real-time to synchronize with unpredictable human movements and maintain a cohesive formation.
Methods
The core of this work is a middleware designed for swarm-driving. This middleware is structured into three key managers:
- Sequence Manager: Orchestrates the overall swarm process, including organizing, maintaining, and dispersing the swarm.
- Message-Processing Manager: Handles communication between robots within the swarm using RF (IEEE 802.15.4 MAC protocol).
- Relative-Location Identification Manager: Enables robots to determine their position relative to each other using IR signal strength.
The system employs a "leader-follower" control method, where one robot (the master) is connected to the user and directs the swarm, while other robots (slaves) follow. To ensure synchronized movement with the user, the master robot broadcasts its driving and user-activity information. Relative positioning is achieved through IR communication, leveraging signal strength to estimate distance and direction. To manage sequential IR communication without hardware conflicts, a token-based system is implemented. The swarm formation is dynamically adjusted based on user activity and driving mode (leading or following).
Datasets
Not discussed in the paper. The evaluation uses a custom hardware setup and experimental scenarios.
Key Insights and Ideas
- Swarm Intelligence for Human Support: Adapting swarm intelligence principles to human-accompanying robots significantly enhances their utility and application scope beyond single-robot limitations.
- Modular Middleware Design: The proposed middleware, with its distinct sequence, message-processing, and relative-location managers, offers a structured and extensible approach to swarm coordination.
- IR for Relative Positioning: Using IR signal strength for relative-location identification proves to be a viable and sufficiently accurate method for enabling swarm formation and synchronization, even with its inherent short-range limitations.
- Token-Based Synchronization: The token mechanism effectively addresses the challenge of time-synchronized IR communication among multiple robots with independent hardware, ensuring reliable data exchange.
- Leader-Follower with User Synchronization: Combining the leader-follower paradigm with real-time broadcasting of user activity allows the swarm to adapt to unpredictable human movements, a critical factor for effective human-robot collaboration.
Key Results
- Relative-Location Accuracy: The relative-location identification using IR communication achieved a Root Mean Square Error (RMSE) of 7.898 cm for distance and 5.8° for direction, deemed sufficient for the main controller.
- Reaction Time: The master robot's reaction time to user activity change was within 1.1 seconds with 90% probability, and slave robots responded within 1.2 seconds with 90% probability. Slaves responded to robot-provided activity information within 0.24 seconds with 90% probability.
- Network Organization Time: The time to organize a swarm network increased with the number of slaves. For 1 to 5 slaves, the minimum required time ranged from 0.6s to 0.8s. The maximum required time also increased, with retry probabilities reaching up to 80% for 5 slaves due to message reception failures.
Limitations and Future Work
- Retry Mechanism: The current retry mechanism for network organization can lead to increased time and retry probabilities, especially with more slaves.
- Relative-Location Accuracy and Acquisition Time: The authors suggest improving the accuracy and reducing the acquisition time of relative-location information to achieve smoother synchronization with user activity.
- Reducing Retries: Future work should focus on decreasing the number of retries during swarm network organization.
- Extended Swarm Sizes: While tested up to 5 slaves, larger swarm sizes might present further challenges.
Summary updated