Abstract
Achieving tightly coupled cooperation in multi-robot systems often requires rigid leader-follower hierarchies that compromise agent independence. This paper proposes a framework within the Hybrid Architecture for Multiple Robots (HAMR) that enables heterogeneous agents to form dynamic teams and synchronize actions using abstract progress metrics. By monitoring internal progress and communicating via completion percentages, robots can cooperate on complex tasks without deep knowledge of their partners' internal mechanics.
Introduction
In the field of multi-robot systems, coordination is typically categorized into collaboration—where agents perform independent tasks—and cooperation—where tasks are mutually dependent. The most challenging form is "tightly coupled" cooperation, such as two robots lifting a heavy object together. Standard approaches often rely on master-slave architectures, which create high mutual dependence and vulnerability to single-point failures. This research explores how independent robots can recognize the need for assistance and synchronize behavior with heterogeneous partners without sacrificing autonomy.
The HAMR Framework
The proposed solution utilizes the Hybrid Architecture for Multiple Robots (HAMR), a three-layer paradigm consisting of a Sequencer, Deliberator, and Coordinator. The core innovation lies in two mechanisms: progress-based recognition and percentage-based synchronization.
Progress-Based Recognition
Instead of relying on a central controller to assign tasks, robots monitor their own progress. If a robot's progress becomes static—for instance, if it attempts to lift a block that is too heavy—it triggers an "impatience" mechanism derived from the ALLIANCE architecture. The robot then notifies the collective, and a partner with the highest utility for that specific task is dynamically assigned to assist.
Percentage-Based Synchronization
To address the challenges of heterogeneity, robots synchronize using abstract completion percentages rather than raw sensor data. This allows robots with different physical capabilities (e.g., different speeds or actuator types) to remain in sync. A leader robot communicates task progress in increments (e.g., 1% to 100%), and the partner adjusts its own actions to match this abstract state, avoiding the "heterogeneity trap" where robots must understand the specific physical parameters of their peers.
Task Structure and Execution
Tasks within the HAMR architecture are defined by specific conditions and actions. This structured approach allows the Sequencer to manage transitions between individual and cooperative behaviors effectively.
As shown in the table above, each task ($T_n$) has a prerequisite condition and a clear end condition. The system defines progress mathematically as a differential measure: $1 - (\text{target position} / \text{current position})$. This simple vector amplitude allows for real-time quantification of success or failure.
Results and Future Work
Initial validation was performed using a simulated heterogeneous robot team, including ActivMedia Pioneer robots and MiniWhegs. The results demonstrated the successful integration of the HAMR architecture and the feasibility of the conceptual framework for dynamic coalition formation.
While promising, the research is in its early stages. Future work will focus on generating precise "expiration times" for task assistance requests and moving from simulation to physical hardware. Testing in noisy, real-world environments will be critical to evaluating the robustness of percentage-based synchronization.
Summary updated