Magnisketch Drone Control

TL;DR

This work tackles the challenge of enabling aerial manipulation for art tasks by introducing Magnasketch, which translates user inputs into magnetic drawings using a low-cost Crazyflie 2.0 and a novel magnetic drawing apparatus. The core contribution is a Model Predictive Control framework that jointly accounts for magnet dynamics and quadrotor motion, coupled with a CAD-driven inertial model and a two-stage onboard/offline control pipeline. Key results show that, while baseline HL control can yield lower raw position error, the magnet-aware LL controller achieves smoother, more aesthetically pleasing drawings with comparable overall performance, enabling full-state control in a lightweight platform. The approach generalizes to both differentiable and non-differentiable shapes and demonstrates feasibility for on-board execution with open-source tooling and hardware, suggesting practical impact for aerial art and manipulation tasks.

Abstract

The use of Unmanned Aerial Vehicles (UAVs) for aerial tasks and environmental manipulation is increasingly desired. This can be demonstrated via art tasks. This paper presents the development of Magnasketch, capable of translating image inputs into art on a magnetic drawing board via a Bitcraze Crazyflie 2.0 quadrotor. Optimal trajectories were generated using a Model Predictive Control (MPC) formulation newly incorporating magnetic force dynamics. A Z-compliant magnetic drawing apparatus was designed for the quadrotor. Experimental results of the novel controller tested against the existing Position High Level Commander showed comparable performance. Although slightly outperformed in terms of error, with average errors of 3.9 cm, 4.4 cm, and 0.5 cm in x, y, and z respectively, the Magnasketch controller produced smoother drawings with the added benefit of full state control.

Figures (28)

• Figure 1: Stippling CrazyFlie
• Figure 2: Flying Calligrapher
• Figure 3: The Crazyflie assumes a body-fixed reference frame, $\large \mathcal{F}_B$, at the geometric center of the drone body. A mass-accurate CAD model was generated to provide key model parameters such as mass distribution and the inertia matrix in the body frame. The system resides in the world frame, $\large \mathcal{F}_W$. The CAD rendering was also used in the mesh cat simulation tool.
• Figure 4: The final manipulator design featured a rough ball-and-socket joint to allow for compliance in z-position, roll, and pitch.
• Figure 5: At a high level, control was split between a precompute phase which generated the full state reference trajectory, and an onboard phase, which utilized Bitcraze commanders to follow these setpoints.