Tails of Great Acrobatics

Quadcopter drones are equipped with four rotors arranged in a square configuration, enabling them to take-off, land, and hover vertically, while also allowing for precise control over movements in any direction. Their utility spans a wide range of applications, from aerial photography and cinematography to search and rescue operations, agricultural monitoring, infrastructure inspection, and even package delivery in urban environments. However, their remarkable capabilities are often constrained by their limited energy efficiency and flight endurance. The relatively short battery life of quadcopter drones restricts their operational duration, hindering tasks that demand extended periods of flight time.

One of the most interesting alternatives to quadcopter drones that you may never have heard of is the tailsitter aircraft. Tailsitters are not new, having first been conceived of in the 1920s by none other than Nikola Tesla. Their name is derived from the fact that they take-off and land on their tails, in an upright position, like a helicopter. When in the air, they can transition into a much more efficient forward flight mode, like a traditional airplane. Leveraging their unique design, they can transition from hovering to forward flight, or somewhere in-between, at any time to perform some amazing acrobatic feats.

Tailsitters are rarely encountered, either in full-size aircraft, or in scaled-down drones, in large part because they are very challenging to control. As far as drones are concerned, some control systems do exist for tailsitters, but due to the difficulties, they tend to focus on calm trajectories and slow transitions between hovering and forward flight. Unfortunately, following that approach fails to take advantage of many of the unique capabilities of a tailsitter.

To unlock the potential of tailsitters as an alternative to quadcopters, a team of engineers at MIT has developed a new control algorithm that can take full advantage of what a tailsitter has to offer. By creating a single, unified model to handle both hovering and forward flight, their system is highly computationally efficient and capable of planning trajectories in real-time. It can plan even very aggressive and complex trajectories that include maneuvers like loops, rolls, and climbing turns.

The technique involves the use of a global dynamics model that handles forward flight, hovering, take-off, and landing. To ensure efficiency of calculations, a technical property known as differential flatness was leveraged in the system. Leveraging this property also allows the aircraft to quickly check, via a simple mathematical function, that a planned trajectory can actually be flown successfully in real-time. This is normally a challenging task because of the complex arrangement of flaps and propellers of a tailsitter, and would otherwise require significant computational resources and time to determine.

MIT’s tailsitter flight control algorithm allows for some very complex maneuvers that match the versatility of quadcopter drones, but with much longer flight times and speed due to the possibility of forward flight. These capabilities open up many new possibilities, like dropping into a collapsed building to rapidly search for survivors while avoiding obstacles with great agility.

A 3D-printed tailsitter airplane was produced to test the team’s control system in a series of experiments. A number of aerobatic maneuvers were demonstrated, like rapid direction changes during high-speed climbs. They also ran some trials in which multiple aircraft performed synchronized movements like loops and sharp turns as they zoomed between a sequence of gates. The researchers note that these maneuvers would not be possible without the real-time trajectory processing made possible by their algorithms.

To date, the system has only been tested in an indoor environment. Moving forward, the team plans to run additional experiments outdoors where environmental factors like wind can adversely affect the aircraft. Once disturbances of this sort can be dealt with, the flight controller would be ready for real-world use. They also intend to look for other applications where their algorithms might have utility in the future.