TL;DR
This paper introduces OptiTrap, a novel computational approach that automatically generates physically feasible and near time-optimal trajectories for acoustic levitation displays. By considering the device's dynamics, OptiTrap enables the creation of more complex and larger mid-air shapes than previously possible, significantly simplifying content creation for this emerging display technology.
What Problem Does This Paper Address?
Acoustic levitation allows for the creation of volumetric displays by rapidly moving small particles in mid-air to form shapes, leveraging the persistence of vision effect. However, a major hurdle has been the difficulty in defining the precise trajectories for the acoustic traps that manipulate these particles. Content creators currently rely on a time-consuming trial-and-error process to determine not only the path of the particle but also the exact timing and positioning of the acoustic traps. This manual process is particularly challenging for complex shapes with sharp edges or significant changes in curvature, often leading to infeasible trajectories that result in particles being ejected from the levitator. OptiTrap aims to automate this process, providing a structured, numerical method to generate these trajectories.
Methods
OptiTrap tackles this problem in two main stages:
- Optimal Path Timing (Challenge 1): The system first determines the optimal timing for a particle to traverse a given reference path (the desired shape). This is formulated as an open-loop optimal control problem. A key innovation here is the development of a non-invertible model for the trap-particle dynamics, which deviates from standard approaches requiring differential flatness. Instead, OptiTrap couples the particle dynamics with a virtual system and uses auxiliary variables to express the forces. This allows for the computation of a timing law that respects the device's capabilities and ensures physical feasibility.
- Trap Trajectory Computation (Challenge 2): Once the optimal timing is established, the system computes the precise positions of the acoustic traps over time. This stage addresses the challenge that the trap's position is not simply along the particle's path due to the nature of acoustic forces. OptiTrap uses the computed forces and timing to solve for the trap positions, employing a numerical method (Powell's dog leg) and introducing regularization to handle potential instabilities.
Datasets
Not discussed in the paper. The paper focuses on a computational approach and its evaluation using synthetically generated shapes.
Key Insights and Ideas
- Non-invertible Dynamics are Key: The paper highlights that the acoustic trap-particle dynamics are not differentially flat, necessitating a departure from standard path-following techniques. This non-invertibility is a core challenge that OptiTrap overcomes.
- Timing is Crucial for Feasibility: The timing of particle movement significantly impacts the required accelerations. OptiTrap's ability to optimize this timing allows for smoother trajectories and avoids infeasible accelerations, especially at sharp corners.
- Trap Position is Not Particle Position: A critical insight is that the acoustic trap's location is not the same as the particle's location. OptiTrap explicitly models and accounts for this displacement, leading to more accurate shape rendering.
- Automation Simplifies Content Creation: By automating the complex process of trajectory generation, OptiTrap democratizes the creation of sophisticated levitation displays, shifting the focus from low-level control to high-level shape design.
Key Results
- Increased Size: OptiTrap enabled up to a 563% increase in the size of rendered objects (e.g., squircle) compared to baseline methods.
- Increased Frequency: Rendering frequencies were improved by up to 150% (e.g., squircle).
- Improved Accuracy: Reconstruction accuracy, measured by normalized RMSE, showed significant improvements, with up to a 77.8% decrease for the squircle and 53.8% for the fish shape.
- High Reliability: OptiTrap achieved a success rate of 10/10 trials for all tested shapes, demonstrating the feasibility of its generated trajectories.
- Near Time-Optimal Performance: The generated trajectories were found to be nearly time-optimal, closely matching device capabilities without exceeding them.
Limitations and Future Work
- Model Accuracy: The paper acknowledges that the accuracy of the trap-particle dynamics model is crucial. Inaccuracies in the hardware or model can lead to distortions. Future work could involve more accurate models or closed-loop control systems.
- Content Creation Tool: OptiTrap is presented as a foundational tool. Future work could integrate it into a more complete content creation pipeline, including artist-friendly interfaces, visualization tools for tricky shapes, and heuristics for tuning parameters like regularization.
- Shape Modification: The current approach optimizes trajectories for a given shape. Future work could explore modifying shapes slightly to improve rendering performance, especially for challenging features.
- Multi-Particle Systems: Extending the approach to handle multiple levitating particles simultaneously is a promising direction, offering greater flexibility and efficiency.
- Other Levitation Modalities: The principles of OptiTrap could be adapted to other levitation techniques, such as optical levitation, with appropriate model adjustments.
- Real-time Closed-Loop Control: Exploring closed-loop approaches, potentially combined with machine learning, could further improve robustness and adaptability to real-world device inaccuracies.
Summary updated