Gallery

Physical human drone interaction (PHDI) with underactuated drones has been widely explored due to its diverse applications from providing guidance to the visually impaired to virtual reality gaming. An opportunity exists in exploiting the decoupled translational and rotational dynamics of over-actuated drones for PHDI.

Watch as an overactuated aircraft gets dragged through the air during human interaction, the aircraft accepting the drag disturbance while maintaining controlled flight.

This work was conducted by Iris Liu and Suvarna Swaraj as a Final Year Research Project in Mechanical and Mechatronics Engineering at the University of Auckland.

Human Drone Physical Interaction

The tops of trees can be hard to access, making trimming and sampling branches difficult and hazardous. What if a drone could do it instead? Here we see a drone extend a manipulator arm to grab a branch swaying below it.

This work was conducted by James Kutia as part of his Doctoral studies in 2019.

Autonomous Branch Sampling with a Drone

If a drone needs to examine a tree, branches can block sensors and tangle in rotors. To avoid this hazard, the drone needs to be able to push away such obstacles. Here we see a branch-spreading system in action, pushing bars apart to peer through. Note that each bar is held by an elastic band, producing a resisting force. In this test, these bands are asymmetric, producing offset forces which the aircraft must account for during its interaction with the branches.

This work was conducted by Pedro Mendes as part of his Doctoral research at the University of Auckland.

Also have a look at the rolled aircraft case here: https://youtu.be/12NPzVsAh68

Separating "Branches" – Asymmetric Loads

If a drone needs to examine a tree, branches can block sensors and tangle in rotors. To avoid this hazard, the drone needs to be able to push away such obstacles. Here we see a branch-spreading system in action, pushing bars apart to peer through (These bars represent branches and have 7kg of resistance through elastic bands). Note that the aircraft is hovering and interacting while rolled over.

This work was conducted by Pedro Mendes as part of his Doctoral research at the University of Auckland.

Also have a look at the asymmetric load case here: https://youtu.be/iyBWLXAwF8g

Separating "Branches" – Rolled

Test flying in the University of Auckland Wind Tunnel involves use of a motion capture array in place of the usual GPS signal. This was an early 2022 test flight to check the functionality of the system prior to research being undertaken later in the year.

What is apparent here, however, with the changing wind speed, is the under-actuated nature of drones: to achieve forward thrust to counter the drag produced by the wind, the entire aircraft is required to rotate. As the wind speed reduces, so does the drag, resulting in the aircraft drifting forwards. The aircraft perceives this drift, and so lessens its angle to compensate.

Flight in the Wind Tunnel

A key task for a drone is to hold position while it takes a photo, or a sample. Being able to reject wind and optimise the aircraft response is key to this. The control methodology is critical to this.

The DTRG Planar Octocopter can be seen here holding position against a 46 km/h wind, with a turbulence intensity of 10%. Using an H-infinity controller halves the root-mean-square (RMS) error from the target hover point compared to the baseline PID control. This work was conducted by Jérémie Bannwarth in 2021 as part of his Doctoral studies at the University of Auckland.

Station Keeping in Wind with H-infinity Control

Modern plantation forest procedures still rely heavily on manual data acquisition in the inventory process, limiting the quantity and quality of the collected data. This limitation in collection performance is often due to the difficulty of traversing the plantation forest environment on foot.

Here we see a small quadcopter overcoming this adverse terrain with autonomous navigation, updating its flightpath with Model Predictive Control. This work was conducted by Tzu-Jui Lin in 2022 as part of his Doctoral Studies at the University of Auckland.

Autonomous navigation through a plantation forest

A conventional drone control system can only respond to the wind once it is disturbed from its target position. What if instead the aircraft can predict how it will continue to move, and plan ahead to counteract the wind, within the limits of its capability?

Here, the DTRG Planar Octocopter is flying in controlled turbulent flow in the University of Auckland Boundary Layer Wind Tunnel, running a Model Predictive Controller (MPC) to do just that. This work was conducted by Caleb Probine and David Yang as a Final Year Research Project in Mechanical and Mechatronics Engineering at the University of Auckland.

Wind Tunnel Testing of Model Predictive Control

Drones offer the ability to measure atmospheric properties in hard-to-reach places. However, the drone itself moves in the wind, distorting wind velocity data. Watch as the small quadcopter is moved about by the turbulent air produced under controlled conditions in the University of Auckland Boundary Layer Wind Tunnel. The velocity as seem by the drone is recorded with a TFI Cobra Probe.

This work was conducted by Lynn May and Ivan Rakich in 2022 as a Final Year Research Project in Mechanical and Mechatronics Engineering at the University of Auckland.

Wind Measurement from a Drone

Subscribe