Abstract
Professional drone cinematography typically requires a coordinated effort between a pilot and a cinematographer. This paper introduces a high-level control system that simplifies this process, allowing a single user to act as both. The system maps complex drone physics and framing constraints into a specialized "Drone Toric Space," enabling real-time path planning, interactive sketching of trajectories, and automated multi-drone coordination. Experimental results demonstrate that the system allows novices to capture professional-grade shots of dynamic targets with high computational efficiency and physical feasibility.
Introduction
In professional filmmaking, capturing dynamic aerial shots is a complex task requiring synchronized expertise in flight and camera control. While consumer drones offer basic "follow-me" modes, they often struggle with maintaining specific cinematographic framing, ensuring physical feasibility of paths, and avoiding collisions in dynamic environments. Furthermore, coordinating multiple drones to provide complementary views without visual interference remains a significant challenge.
Methods
The proposed system architecture is built around a novel mathematical representation called the Drone Toric Space (DTS). The DTS re-parameterizes a spindle torus surface to embed safety distances and physical gimbal limits directly into the coordinate system. This ensures that the search space for valid shots is pre-filtered for safety and feasibility.
The system employs a multi-layered control scheme:
- Real-time Path Planning: Using a visibility-aware roadmap and A* search in a 4D $\tau$-space, the system finds paths that optimize for visual properties rather than just Euclidean distance.
- Smooth Trajectories: To ensure stable flight, the system generates $C^4$ continuous trajectories (piecewise degree-5 polynomials), accounting for "snap" to minimize jerky movements.
- Interactive Control: A sketching interface allows users to draw 2D paths on a screen, which are then projected and optimized into 3D flight paths.
- Multi-Drone Coordination: A master/slave hierarchy uses "min-conflict" optimization to assign distinct framing instances to multiple drones while preventing them from entering each other's field of view.
Evaluation and Results
The system was evaluated in both simulated Unity 5 environments and a real-world lab setup using a Vicon motion capture system for high-precision tracking. In a user study with 12 novices, the proposed tools significantly outperformed manual controllers, enabling users to achieve fluid, precisely framed shots that were previously impossible for them.
Key findings include:
- Efficiency: Path planning algorithms consistently run in under 200ms, supporting real-time interaction.
- Smoothness: Quantitative analysis confirmed that DTS-based planning results in smoother footage with fewer variations in target framing compared to standard Euclidean methods.
Conclusion
By abstracting low-level flight controls into high-level cinematographic directives, this system empowers solo creators to produce professional-quality aerial footage. While currently limited by reactive planning and hardware constraints of non-stabilized drones, future work aims to integrate predictive motion schemes and automated real-time editing for even more complex multi-drone behaviors.
Summary updated