Robotics

   Introduction

We live in a world where everything can be controlled and operated automatically, but there are still a few important sectors in our country where automation has not been adopted or not been put to a full-fledged use, perhaps because of several reasons one such reason is cost. One such field is that of agriculture. Agriculture has been one of the primary occupations of man since early civilizations and even today manual interventions in farming are inevitable.

The idea of robotic agriculture (agricultural environments serviced by smart machines) is not a new one. Many engineers have developed driverless tractors in the past but they have not been successful as they did not have the ability to embrace the complexity of the real world. Most of them assumed an industrial style of farming where everything was known before hand and the machines could work entirely in predefined ways – much like a production line. The approach is now to develop smarter machines that are intelligent enough to work in an unmodified or semi natural environment. These machines do not have to be intelligent in the way we see people as intelligent but must exhibit sensible behavior in recognised contexts. In this way they should have enough intelligence embedded within them to behave sensibly for long periods of time, unattended, in a semi-natural environment, whilst carrying out a useful task. One way of understanding the complexity has been to identify what people do in certain situations and decompose the actions into machine control is called behavioral robotics

The approach of treating crop and soil selectively according to their needs by small autonomous machines is the natural next step in the development of Precision Farming (PF) as it reduces the field scale right down to the individual plant or Phytotechnology (Shibusawa 1996). One simple definition of PF is doing the right thing in the right place at the right time with the right amount. This definition not only applies to robotic agriculture (RA) and Phytotechnology but it also implies a level of automation inherent in the machines. Automatic sensing and control (on-the-go) for each task is also important and many research papers have shown that these systems are feasible but most are too slow, and hence not economically viable, to be operated on a manned tractor. Once these systems are mounted on an autonomous vehicle, they may well suddenly become commercially viable.

By taking a systems approach, in which we consider a system in terms of its actions, interactions and implications, we can develop a new mechanization system that collectively deals with all the crop’s agronomic needs in a better way. To do this we must stop defining plant care in terms of the current mechanisation but in terms of what the plant needs. When we have defined the actual plant requirements we are then free to design a better way of dealing with them.

The environmental implications would seem good. Minimised inputs to reduce waste and pollution, controlled biodiversity by retaining non-competitive weeds, more intelligent physical methods replacing chemical solutions are all examples of how Phytotechnology could benefit the environment over traditional methods. Economic factors include lower labour costs (a significant saving if they can be made fully autonomous), incremental investment in, perhaps, a small machine each year, rather than a single large machine every 5 years. These small vehicles could be assembled from existing mass produced components such as car parts without the need for specialized design and tooling. Consideration of social aspects shows that the public are ready for small intelligent machines to be used in food production, by the level of interest shown by the media and when being demonstrated. Insurance and liability will be a lot easier with smaller autonomous machines.

Most of the current machinery is very weather dependant. Tractors cannot drive on soil when it is wet, sprayers cannot work in high winds etc. Perhaps it will be possible to develop smaller, less intrusive machinery that can allow more tasks to be carried out in marginal conditions. An example might be an autonomous seeder that could function well, while the soil is still wet in the springtime, provided that the soil engagement mechanism is suitable arranged. This would allow the seeds to be planted when optimal for the crop and not be limited by the soil’s ability to support the tractor.

Safety is another important factor. Any autonomous vehicle is going to go wrong at some time and the chance of catastrophic failure should be minimized within the design process. A small light vehicle is inherently safer than a large one. Redundant, self checking systems should be built into the system architecture to allow graceful degradation. The vehicle should be in continual communication with the base station, giving data about current conditions and contexts this approach may not be economically justifiable in many broad acre crops but will certainly be more attractive in high value crops where a smart machine can replace expensive repetitive labour. If this approach were taken, it would appear that the crop production cycle could be reduced to three stages: Seeding, Plant care and (selective) harvesting.