Home / GPS&GIS / GPS for Precision Farming

GPS for Precision Farming

Precision farming is a method of crop management by which  areas of land within a field may be managed with different levels of input depending upon the yield potential of the crop in  that particular area of land. The benefits of so doing are two  fold:

− the cost of producing the crop in that area can be  reduced;

− the risk of environmental pollution from  agrochemicals applied at levels greater than those required by the crop can be reduced.

Precision farming is an integrated agricultural management  system incorporating several technologies. The technological  tools often include the global positioning system GPS,  geographical information system GIS, remote sensing, yield  monitor and variable rate technology.

The paper talks about the use of GPS to support agricultural  vehicle guidance. Equipment for this purpose consists on a  yield monitor installed: the system supports human guide by  means of a display mapping with a GIS the exact direction  produced by GPS receiver put on vehicle top: the driver  follows it to cover in an optimal path the full field.

GPS receivers for this applications require, not only an high accuracy to ensure the reduction of input products, but even an  easy and immediate way of use for farmers; without forgetting low costs. Obviously the technology to achieve high precision still exists but it is too expensive and difficult to use for not skilled people. Survey modality usually adopted in agricultural applications is real time kinematic positioning, DGPS RTK, which enable tohave a good accuracy by means of corrections received. In this experimentation the aim is to obtain a sub-metric accuracy using low cost receivers, which can provide only point positioning. These receivers have been developed for maritime navigation purposes; our aim is their optimization in order to apply them for land navigation in particular for farming activities. Some tests using these receivers were carried out, but results were not satisfying and probably the reason has to be assigned to the implementation of a Kalman filtering inside the receiver software. This is the starting point for a new project, at the moment still in progress, which aim to develop a new  algorithm based on Kalman filter. Its purpose is to improve low cost receiver outputs in order to optimize trajectories and to reach needed accuracy in vehicle positioning during agricultural  activities.

TRIAL AND ERROR

1 Instruments and tests

Experimentation has been carried out using Leica Geosystems  instruments; in particular the low cost receiver discussed in the paper is the TruRover Leica. Its mainly features are: it is an antenna-receiver integrated instrument, it has a 5 Hz tracking time, the report is in the NMEA string format, it cannot neither store positions nor show them in real time, it requires a computer to view NMEA data stream. TruRover performances were compared with geodetic receiver one, which are considerably better, so they are the perfect comparison condition to estimate Trurover positioning quality.

Geodetic receiver used is the GX1230 Leica, able to receive double frequency (both code and phase). Both static and kinematic tests were performed, simulating the typical behaviour of an agricultural vehicle (straight and parallel trajectories with reduced velocity, such as 20÷40 km/h)

and using, at the same time, the two different kinds of GPS receivers described above. At the top of the vehicle, both TruRover and geodetic antenna, connected to the receiver, were placed at a distance of 50 cm. Three static stops with 20 minutes time length were performed, spaced with two steps in motion. Geodetic receiver were set with a 1 second tracking time and a cut off angle of 10 degree. Tests length were about two hours. Another geodetic receiver were placed for a single point positioning and used as the Master station for the following data processing.

2. Data processing

Master station coordinates were determined by means of a static processing in relation to two different GPS permanent station in order to check result: one placed in Modena, where tests have

been carried out, led by INGV and the other located near Bologna, led by ASI Telespazio.

TruRover NMEA data already contain coordinates and Visual GPS software has been utilized to show and store them. These positions have been compared to data stored by double frequency receiver during kinematic tests. These data were utilized to estimate the exact trajectory, which was estimated by the postprocessing in kinematic differential modality. Software for data processing was Leica Geo Office. To be honest this trajectory is not exact because even kinematic postprocessing data have some errors; however this modality has a centimetric accuracy, better than the required from agricultural applications one so it is not a mistake to consider this track as an exact one. TruRover track and the exact one are not yet comparable because 50 cm shift still exists: a kind of overlap has been done by means of setting vehicle motion direction thanks to postprocessed trajectory.

3 Results analysis

The results of the comparison between TruRover track and double frequency receiver one are not satisfying; indeed receivers utilized in experiments show some problems in curves, where the estimated track is larger than the exact one. This bad performance may be due to the presence of a Kalman filter inside the system, that is not optimized for the specific application. Probably at each epoch this filter uses previous estimated positions in order to anticipate the future one on a constant velocity, linear trajectory assumption. In that way when vehicle curves the filter understand it as a mistake and modify the position; this behaviour causes a delay in curving and consequently a shift in positioning.

Higher precision for agricultural applications is not required in curves but in straight directions, where farmers make their main activities on yield. However curves have a great importance mainly at their end because there it is necessary for the vehicle trajectory to be parallel to the previous one. The main reason for that is to economize input products spread about field. Kinematic trajectory is considered the exact one, the reference for a comparison between pseudo-range and kinematic tracks.

The results show distances greater than 1 meter (the target aimed) but always inside the method precision (10 meters). Statistical parameters, as means and standard deviations,  confirm the same things. Table 2 and 3 relate these statistical  valuers. At the beginning the idea was that Kalman filter needs a period of assessment time to work better; on the contrary,  with the elapsed time the differences increase with a worrying time drift.

DEVELOPMENT OF A NEW ALGORITHM BASED

ON KALMAN FILTERING

The reason for problems in curve is probably the presence of a Kalman filtering inside TruRover, not especially studied for farming applications. Thereof the need of trying a kind of TruRover performances improvement pursued by means of the development and the implementation of a new algorithm based on Kalman filtering and, at the same time, optimized for agricultural requirements.

The first problem was the choice of the process modelling to put in Kalman equations. In particular two trials have been done and described in the following: the constant velocity model and the constant acceleration model. Before the models description, it will be shortly illustrated Kalman filter principles.

CONCLUSIONS

The above described problems are a great problem for precision  farming because bad tracks in the field cause wastes of material, without considering economical and environmental impacts. So that, starting from the analyses of the previous results and taking into account the typical user requirements, a preliminary design for the new algorithm based on Kalman filtering has been done. The idea underlying the new navigation system is to implement a simplified version of the so called adaptive Kalman filtering; the filter takes into account both the typical behaviour of an agricultural vehicle and the a priori knowledge of the planned track and works continuously testing alternate hypotheses in predicting the track. The new Kalman algorithm should both eliminate drifts in curves and occasional spikes in satellite configuration changes. This research project is still in progress; at the moment we have implemented the new algorithm which consists on a double filtering using the constant velocity model in straight trajectories and the constant acceleration model for curve tracks.

Problems during algorithm testing were mainly the lack of raw data, in fact TruRover NMEA reports are still filtered and there is not the possibility to remove the previous filter implemented inside the receiver and it is not mathematically correct utilizing them for another filtering. For

this reason data inputs for new algorithm have been provided from double frequency receiver without post-processing (raw data really as they have been stored). Results confirm the importance to adopt a model based on acceleration in curve, but at the same time it is necessary looking at these results in a critical way because they are outputs originated from inputs  better than Trurover data. In the tests the attention will be mainly focused on variables which have a great importance in the model and parameters choice, such as process covariance and measurement noise. Next steps will be two-fold:

− trying to vary covariance weighs both in system noise matrix and in measurement noise matrix;

− test double filtering with raw data not yet filtered and tracked by a low cost and single frequency receiver, showing located spikes.

The purpose to improve TruRover performances and to optimize them for precision farming is challenging, especially having at our disposal only raw data. Other possible solutions are:

− connecting an odometer and a steering wheel to the system, integrated with the GPS receiver, which supports human vehicle guide. It could be the input to choose, at the right time, the best process model to adopt inside Kalman filter (constant velocity or constant acceleration model).

− utilizing differential positioning, DGPS, improving coordinates thanks to corrections received from a Master station close to the field.

(Source – http://www.isprs.org/proceedings/XXXVI/5-C55/papers/biagi_ludovico_1.pdf)

Leave a Reply

Your email address will not be published. Required fields are marked *

*