Collaborative Task Planning for PCB-Driven Magnetic Microrobots Using Deep Reinforcement Learning

Magnetic field-driven microrobots hold great promise for medical applications due to their ability to safely navigate within complex environments. Traditional global magnetic field approaches are typically limited to controlling single microrobots, making multimicrorobot coordination a challenge. To overcome such challenge, we develop a novel microrobot system driven by a 12 × 12 microcoil array on a printed circuit board, which generates localized magnetic fields for independent control of multiple microrobots. This system enables autonomous multitask planning under complex obstacle constraints. Using deep reinforcement learning enhanced by A* algorithm guidance, we facilitate efficient and dynamic collaborative task planning. The proposed approach improves adaptability and coordination between microrobots in environments with multiple obstacles. Experimental results, including tests in dynamic complex environment with multimicrorobots, demonstrate that our system effectively achieves coordinated, interference-free paths. This work offers a significant advancement in dynamic task planning and coordination of multimicrorobots, providing a powerful solution for complex collaborative tasks such as autonomous and efficient reagent detection and drug fusion in medical treatment.