Choose an application

• Reference Manager
• EndNote
• RefWorks (Direct export to RefWorks)

Unmanned Aerial Vehicles (UAV) are currently flying over fruit orchards performing water stress and disease detection, but with tree-level precision. If they were to fly through the fruit orchard corridors and use sidewards looking cameras, individual leaves and fruit could be inspected at close-range and high spatial resolution. This would enable earlier and more accurate disease or water stress detection, but also harvest yield estimations and individual fruit quality assessments in which size, shape, colour, texture and fruit damage from hail, birds or other small defects could be assessed. This data could greatly benefit the fruit grower and allow specific adaptations in the cultivation process and could even be used by autonomous ground vehicles which could perform mechanical tasks such as pruning, spraying and harvesting adapted to each individual tree or fruit. In order for the UAV close-range fruit inspection to be (commercially) viable, two main challenges can be identified: (i) the design and construction of a suitable rotorcraft UAV for narrow corridors and (ii) the autonomous navigation of this rotorcraft UAV through the fruit tree corridors. The narrow corridor, through which the rotorcraft UAV has to fly, limits its width and thus, for conventional helicopters and multicopters, the size of its rotor/propellers, total mass it can carry and flight endurance. A new type of multicopter is conceived and defined: the compound multicopter. A compound multicopter is defined as a multicopter that uses one set of propulsion units to lift most of the multicopter's weight while another set controls its motion. A compound multicopter can be "stretched" lengthwise and combine large (efficient) lift propellers with small (fast response) control propellers, resulting in higher payload capacity, longer flight endurance for a given width or a smaller width for the same total mass. Several narrow corridor compound configurations were conceived and their corresponding prototypes constructed and flight tested to find the best one upon which the final prototype is based. The various compound prototypes performed several dozen flight tests in fruit orchards capturing various types of fruit cultivated in different fruit systems. Analysis of the (in-flight) acquired fruit images, which have spatial accuracies down to a few millimetres, proves they are indeed suitable for automated individual fruit detection and inspection. Two tools were developed to aid the design of a narrow corridor compound multicopter and, for comparison purposes, for two types of conventional multicopter: quadcopter and co-axial octocopter. The first tool is a performance estimation tool and the second tool is an automated component-based detailed conceptual design tool. All tools were validated through static motor-propeller testing and endurance flight tests using the various (compound) prototypes. The robustness to outdoor atmospheric disturbances (wind and gusts) of both conventional and various compound multicopters was evaluated by quantifying their agility and manoeuvrability through an experimental, simple and fast new test procedure. A modular and generic constraint-based task specification approach and UAV flight control system architecture was developed consisting of a low-level flight controller and a high-level navigation and data-rich sensor processing module. iTaSC was chosen as the preferred constraint-based method for the navigation module. The architecture is capable of simultaneously performing several tasks: human-machine shared control, (dynamic) obstacle avoidance and (visual) object (marker or corridor) and GPS-trajectory tracking in unknown environments, both in- and outdoor, employing only on-board sensors and (limited) computational power. The flight control system architecture provides fast, reactive behaviour and a more intuitive feedback, while retaining as much direct user control freedom as possible. The architecture and set-up were extensively and experimentally validated both in- and outdoor. The indoor scenario was performed in a GPS-denied environment and assumed a human operator as primary pilot who has to fly the UAV through a narrow corridor. The pilot has shared control and visual feedback of the UAV through a forward-mounted camera that streams the video feed in real-time. The flight controller assists the pilot by performing obstacle avoidance and either visual object (marker) tracking or corridor tracking. The outdoor scenario was performed in a fruit orchard and assumed the human operator as back-up safety pilot who only monitors the autonomous mission, but can overrule the high-level navigation module's commands through shared control in case of emergency. The flight controller performs several tasks autonomously such as sonar- and camera-based corridor tracking, GPS-trajectory tracking and (sonar-based) obstacle avoidance. The experiments showed that the UAV can autonomously enter, fly through and exit a fruit orchard corridor while allowing immediate intervention by a safety pilot. Autonomous flight tests through fruit orchard corridors without collisions were successfully completed in more than 90\% of the attempts even in the presence of (low) wind and turbulence.