Aerial Manipulation for Canopy Sampling

Abstract

The uptake of unmanned aerial vehicle technology within the forestry industry has accelerated in recent years, addressing remote sensing challenges such as forest yield and health monitoring. Nevertheless, forest management cannot be realised through remote sensing alone; physical canopy branch samples are required to test for diseases, analyse nutrient levels and conduct genetics research. Motivated by the advent of aerial manipulation research, this work proposes physical sampling of canopy branches with an unmanned aerial vehicle. This would replace conventional methods such as tree climbing or firearm dislodgement, which are labour intensive, time consuming and expensive. Development of a novel Canopy Sampling Aerial Manipulator (CSAM) is presented from inception, with design features motivated by unique application-related challenges. A 6 degree-of-freedom manipulator is longitudinally offset from the aircraft centre to avoid rotor wake disturbances to branches and augment end effector reach, facilitating sampling operations from above and beside candidate branches. Cutting and retaining branch samples is achieved with a specialised end effector. Flight control is implemented using dedicated hardware and interfaced with an onboard computer for high-level manipulation tasks. Operations are restricted to a motion capture environment, avoiding onboard perception development to focus on other important aspects of the canopy sampling operation. Aerial manipulator dynamics are modelled and characterised using existing methods, including disturbances imposed by the manipulator movement. This yields accurate simulation of free-flight scenarios to support experimental testing. Physical interaction with branches is a novel aspect of canopy sampling, therefore the model is expanded to consider additional disturbance effects from branch coupling. Comparisons between experimental and simulated flights validate the model and give insights to aircraft station-keeping behaviour during the coupling. Automated branch capture methodology is developed, for stationary and swaying branches. An extended Kalman filter is formulated to estimate branch sway parameters from position measurements. Capture is attempted by the manipulator along the major principal sway direction, while the aircraft undergoes station-keeping. Performance is validated in simulation and experimentally using the motion capture environment, where influences such as station-keeping perturbations and deviations in sway motion are also investigated. This work presents a significant step towards the aim of autonomous canopy sampling, with proposed future work addressing the transition to outdoor operations with implementation of onboard vision.