110 km/h Vehicle Landing: University of Sherbrooke Breaks a High-Speed Multirotor Landing Barrier

Can a multirotor land stably on the roof of a vehicle moving at 110 km/h, under unstable airflow and strong turbulence? Under strong wind, turbulence, platform vibration, and large relative speed, the traditional “slow, level, gentle touch” landing method often fails.

In 2025, the Createk team at the University of Sherbrooke published research in the Journal of Field Robotics proposing a combined approach using friction shock absorbers (FSA) and reverse thrust (RVT). At touchdown, the friction mechanism dissipates impact energy, while motor reversal provides downward holding force. Together, they significantly expand attitude and velocity tolerance and enable stable multirotor landing on high-speed moving platforms.

Video source: https://www.youtube.com/watch?v=tTUVr1Ogag0

Why Is Landing Hard for Conventional Multirotors?

In high-speed and disturbed environments, conventional multirotor landing becomes difficult for several reasons:

• Attitude conflict: a large pitch angle is needed to counter drag, but conventional landing gear requires near-level contact, otherwise bouncing or rollover may occur.
• Bounce and sliding: rigid or elastic landing gear stores impact energy and then releases it during rebound. Under turbulence and relative wind, this can lead to sliding or overturning.
• Strict timing: gusts, estimation errors, and platform motion make “touching down at the exact correct moment” extremely difficult.

Research Idea

The goal is to enlarge the set of states that allow safe touchdown under strong drag, turbulence, and random platform acceleration. This relaxes the requirements on perception accuracy, control bandwidth, and timing, allowing faster descent while still landing safely.

Fast Descent

The team increased baseline descent speed to about 3-4 m/s. This “fast approach, short exposure” strategy reduces the time spent in the most turbulent airflow above the vehicle roof. Reliable dissipation of touchdown energy and resistance to sliding are then handled by the friction shock absorbers and reverse thrust.

The FSA converts vertical kinetic energy into heat and avoids rebound, while RVT increases normal force and friction by pressing the aircraft down at touchdown. In this way, dependence on high-bandwidth sensing and precise control is partly transferred to a designed physical energy-dissipation mechanism.

Late Rapid Leveling

During vehicle following, the UAV needs a large pitch angle to resist drag. If it levels too early, horizontal speed mismatch with the vehicle increases. If it levels too late, the propellers may contact the platform first. The strategy is to perform a rapid leveling maneuver within a very short window before touchdown, so contact occurs in the middle of the leveling process. Remaining angle and angular-rate errors are absorbed by FSA and RVT.

Expanding the Touchdown Envelope

The joint-friction FSA replaces elastic energy storage with frictional energy dissipation. Each landing leg uses friction disks that rotate relative to each other at touchdown, converting vertical impact energy directly into heat and minimizing rebound. Rubber feet increase friction, and RVT presses the aircraft onto the platform, suppressing sliding and rollover.

Platform and Implementation

To implement the strategy, the team designed the DART UAV system. The complete aircraft with landing gear weighs about 2.4 kg, has a thrust-to-weight ratio of about 6.5, and reached a measured top speed of about 126 km/h. The propulsion system supports fast motor reversal for RVT, while customized FSA landing gear dissipates energy and improves landing stability.

Relative positioning is provided by onboard RTK and two RTK beacons on the vehicle, including heading information. The UAV and beacons communicate through ESP32 peer-to-peer links at about 20 Hz, with RFD900+ for telemetry and the ground station forwarding RTK corrections.

Near-ground aerodynamic measurements showed a significant boundary layer above the vehicle roof. Air speed close to the surface was reduced by up to about 32% compared with free-stream speed, with a boundary-layer thickness of about 42 cm. This explains why the UAV did not slide or get blown away as much as expected at high vehicle speeds, and the effect was included in the simulation model.

Experimental Verification

The team conducted outdoor tests on 200 m and 400 m straight road sections at vehicle speeds from 10 to 110 km/h. A total of 38 vehicle-following landings were completed, all successfully.

At touchdown, vertical speed was about 1.4-3.0 m/s, relative horizontal speed about -0.9 to 1.1 m/s, pitch angle about -41° to 0°, and pitch rate about 0-240°/s. After touchdown, the aircraft did not rebound and instead stayed attached to the platform. Even at 110 km/h, sliding on the platform was about 2 cm or less.

Monte Carlo Simulation

The team further used Monte Carlo simulation to sample wind field, vehicle acceleration, state-estimation error, and boundary-layer uncertainty. The tests covered descent speeds of 1, 2, and 3 m/s and vehicle speeds of 6-34 m/s, roughly 20-120 km/h, for a total of 90,000 simulations.

Rigid landing gear showed lower success rate at higher descent speeds because of rebound and a smaller feasible envelope. FSA + RVT, by contrast, improved success rate as descent speed increased from 1 to 3 m/s and maintained about 80% success near 103 km/h under the best descent-speed setting.

Resources

• Paper: Friction Shock Absorbers and Reverse Thrust for Fast Multirotor Landing on High-Speed Vehicles
• Paper link: https://doi.org/10.1002/rob.70069