This research proposes Gen-CoDrone, a framework extending the CoDrone platform for heterogeneous aerial-ground robot coordination. It introduces a hierarchical architecture comprising a heterogeneous abstraction layer, distributed capability-aware task allocation using multi-agent reinforcement learning, and optimized coordination protocols. The validation plan includes Gazebo/PyBullet simulations and real-world trials with CoDrone EDU and TurtleBot3 units. The framework targets 35-50% improvement in multi-modal missions over homogeneous approaches while operating within educational hardware constraints.
Key findings
Heterogeneous robot abstraction layer (HRAL) unifies CoDrone aerial units with ground platforms like TurtleBot3.
Lightweight capability-aware task allocation algorithm using graph attention networks optimized for resource-constrained educational hardware.
Expected 35-50% performance improvement in multi-modal mission scenarios compared to homogeneous approaches.
Open-source simulation and evaluation framework with standardized benchmarks for heterogeneous educational robot systems.
Limitations & open questions
Focuses on educational-grade hardware with limited computational resources compared to industrial platforms.
Proposed evaluation limited to specific mission scenarios including search-and-rescue and area coverage.
Framework assumes specific communication protocols that may require adaptation for other robot combinations.