Pesticides Research, 97 – Autonomous Weeder for Christmas Trees – Basic Development and Tests

Autonomous Weeder for Christmas Trees – Basic Development and Tests

2 Previous studies and development work

2.1 Recent developments in field robot technology
2.2 Specific control related issues

The present methods of Christmas tree cultivation were described in details in the preliminary feasibility study (Have et al. 2002). The minimum requirement to weeding is determined by: 1) Weed competition on water and nutrients and 2) Damage to the tree branches because of wear against woody weeds. The weed competition was found to be insignificant when weeds were removed within a radius of approximately 40 cm around each tree, when they are small. With increasing tree size weeding is needed in a larger diameter, while the need for weeding close to the trunk is reduced because of low light intensity. Damage of branches can occur during the sensitive sprouting period if weeds are close enough to rub the growing points when moved by the wind. Weeds outside these areas were considered to be harmless or even beneficial, as they may improve biodiversity and have sheltering effects to small trees as well. However, during spring all weeds should be removed to allow heat radiation from the ground to reduce frost damage of the sensitive growing points.

The idea and overall concepts of a small, autonomous weeding machines was outlined, including suggestions of operation patterns, physical designs of the vehicle and the system architecture with sensors, databases, navigation system, and actuators. Also the likely costs of using such a machine were estimated and compared to current contractor rates.

This study led to the conclusion, that it is realistic to develop an autonomous Christmas tree weeder from a technical point of view, and also that it is possible to fulfil the aim of weeding only part of the area. In addition the machine was considered suitable for collection of tree specific data concerning condition and development for use in plantation management and product documentation.

It was also estimated in the study that such a concept would have significantly lower environmental effects than present technologies (spraying and mechanical weeding). In the future it could be of considerable interest in other special cultures such as fruit and berry orchards and plantations.

The control concept was based on centimetre precision of the geo-position of each tree as well as the current position of the ACW by means of a RTK GPS (Real Time Kinematic Global Positioning System). The positions of the trees could be measured automatically during machine planting or surveyed-in later on. The resulting tree map is the basis for generating a suitable route plan for navigation over the entire lifetime of the plantation.

As a weed control tool, a rotor cutter similar to those used in hand held grass cutters (“Strimmers”) was proposed, because it requires less energy than tillage tools, while the effect on weeds is nearly the same as by tillage.

Two platform concepts were proposed: 1) A small, stable machine able to move between the trees. This gives a low mobility platform and a high mobility cutter. (The machine drives in a straight line while the cutter moves in and out) 2) A portal machine with a set of wheels spanning the row and a working tool on either side of the row. The first of these was found to be the simplest and best suited to the purpose.

2.1 Recent developments in field robot technology

Autonomous vehicles are widely used in industrial production and warehouses, where controlled environments can be guaranteed. Autonomous vehicles for use in open environments are much more difficult to develop and this work is at an early stage. No autonomous field machines have yet been marketed for professional use; but a few experimental vehicles have been developed. An overview of the research and development in the area is provided in the feasibility study (Have et al. 2002).

The work on field machine automation started long ago, (the first radio controlled tractor was developed in the 1920s) but it was only with access to the high precision RTK GPS and other advanced sensor systems that it became practically realistic to build autonomous field machines (Wilson, 2000; Hollingum, 1999). Development work is now being undertaken in several places, notably Japan and America. A few publications have been published on fully autonomous vehicles, such as the Demeter system for harvesting equipped with a video camera and a GPS for navigation (Pilarski et al. 2002), as well as semi-autonomous vehicles.

During recent years, many researchers have started to develop more rational and adaptable vehicles based on a behavioural approach (Blackmore and Fountas et al. 2004; Bak and Jakobsen 2003). Much effort has been put in to developing control systems and architectures to deal with the complexity of the interaction between robots and the semi-natural environment.

2.2 Specific control related issues

The approach, and hence the system architecture, that has gained increased attention is the hybrid system. This approach combines both deterministic control and reactive behavior. The deterministic control is hierarchical and is used for pre-planned operation in a structured and known environment (e.g. a field), while the reactive behavior is used to respond directly to events and is used for unstructured or unknown environments and conditions (e.g. a tree fallen over in a field).

The advantage of reactive behavior is that it can deal with uncertainty in perception. In contrast, a hybrid system combines reactive and goal-oriented control. These hybrid approaches, based on behavior are therefore the centre of focus for many researchers and there have been many control architectures proposed (e.g. Yavuz and Bradshaw, 2002). Recent work at KVL has demonstrated a deterministic tractor capable of carrying out field operations and a fully reactive research robot capable of exploring unknown areas (Blackmore and Griepentrog et al. 2004). A hybrid system that uses the advantages of both techniques is needed to make a practical system.