On the Road Again

Quadcopter drones are being utilized more frequently than ever in various industries due to their versatility and maneuverability. These small, unmanned aerial vehicles are equipped with four rotors, making them capable of hovering in place, moving in any direction, and even performing acrobatic maneuvers.

One of the primary applications of quadcopter drones is in aerial photography and videography. With their ability to fly at different altitudes and angles, these drones can capture stunning footage from above, providing a unique perspective that was previously impossible to achieve. This technology is used in real estate, film production, sports broadcasting, and even in the surveying and mapping of terrain.

The list of additional application areas for drones is long and diverse, ranging from agriculture and delivery to security and search and rescue. But to expand their utility even further, and enable new use cases, these vehicles need to learn a few new tricks. One area where drones struggle at present is landing on surfaces in motion, especially when those surfaces are not flat and horizontal.

Drones capable of this feat would be able to, for example, land on a moving car. This could allow them to periodically be deployed to capture detailed analyses of the environment, then hop back on the car to hitch a ride and conserve their battery until they are in position for their next mission.

It is exactly this functionality that a team of engineers from Shanghai Jiao Tong University sought to develop. Their interest in this problem stems from their observation that in addition to cars, many structures, like buildings and bridges, also have inclined surfaces that are difficult to land on.

To enable new use cases that require landing on such surfaces, they have fitted a quadcopter named Hitchhiker with self-sealing suction cups in a wheel-like configuration. This arrangement of suction cups allows the drone to get a tight grip on many types of surfaces even if its approach to the target is not quite perfect.

But that only covers the final moments of the landing — the drone still needs to get very close to the target before it can grab hold. For that purpose, a trajectory planning algorithm was developed. This algorithm employs a two-stage tracking approach that considers both the drone’s position and attitude to adjust the individual thrust of each rotor. As it stands today, this tracking algorithm relies on an external camera which heavily restricts the actual use cases Hitchhiker can be applied to, however, the researchers are exploring onboard sensing options for a future version of the system.

A number of experiments were conducted to determine how well the system would perform under realistic conditions. A mock car was created with a panel on the back that could be adjusted to different degrees of incline. As this car traveled along a straight path, Hitchhiker was demonstrated swooping in and successfully landing by attaching its suction cups. It was successful in over 70 percent of cases, and at a wide range of surface inclines. It should be noted, however, that the maximum speed of the car was about 2.4 miles per hour, so it is not exactly representative of a real world scenario in which a drone lands on a moving vehicle.

There is still plenty of room for improvement aside from the relatively low speeds that the drone can presently land at. It was also noted that the cluster of suction cups that are needed for smooth landings create drag and other airflow disturbances in flight, which can hinder normal flight operations.

But in spite of these shortcomings, the team is learning a lot as they continue to develop the system. They have found that, somewhat surprisingly, the drone has more successful landing attempts when the vehicle is moving backwards, for example. After addressing the known shortcomings, and improving their methods from the knowledge they have gained during development, their platform may eventually be ready for commercialization, which is the researchers’ ultimate goal for this work.