Efficient Model Learning and Adaptive Tracking Control of Magnetic Micro-Robots for Non-Contact Manipulation

Abstract

Magnetic microrobots can be navigated by an external magnetic field to autonomously move within living organisms with complex and unstructured environments. Potential applications include drug delivery, diagnostics, and therapeutic interventions. Existing techniques commonly impart magnetic properties to the target object,or drive the robot to contact and then manipulate the object, both probably inducing physical damage. This paper considers a non-contact formulation, where the robot spins to generate a repulsive field to push the object without physical contact. Under such a formulation, the main challenge is that the motion model between the input of the magnetic field and the output velocity of the target object is commonly unknown and difficult to analyze. To deal with it, this paper proposes a data-driven-based solution. A neural network is constructed to efficiently estimate the motion model. Then, an approximate model-based optimal control scheme is developed to push the object to track a time-varying trajectory, maintaining the non-contact with distance constraints. Furthermore, a straightforward planner is introduced to assess the adaptability of non-contact manipulation in a cluttered unstructured environment. Experimental results are presented to show the tracking and navigation performance of the proposed scheme.

Figures (6)

• Figure 1: Illustration diagram about magnetic robot performing non-contact manipulation on the target object. The rotating magnetic field generated by the electromagnetic coil propels the robot to roll, thereby inducing fluid motion to drive the target object.
• Figure 2: Overview of the proposed non-contact manipulation scheme. (a) One-dimensional schematic diagram of non-contact pushing. (b) Two-dimensional motion model employing a local coordinate system, where the normal and the tangential are coupled. (c) Control structure of non-contact manipulation. The controller receives the reference trajectory from the planner and the estimation model based on the neural network, then gives the control input through optimization and online updating.
• Figure 3: System setup. The three-axis Helmholz coil generates a rotating magnetic field, while a camera is connected to an inverted microscope to capture images. The magnetic field inputs are calculated on the PC side.
• Figure 4: Validation of the data efficiency of local coordinate projection and decoupling method
• Figure 5: Trajectory tracking of "ICRA". The green circle represents the target object, the red circle represents the magnetic-driven robot, the blue line represents the reference trajectory, and the orange line represents the actual trajectory.