Development of a Sensing System to Assess Automatic Steering Systems

Transcription

1 Development of a Sensing System to Assess Automatic Steering Systems M.Sc. S. Groth, Prof. Dr. H.W. Griepentrog, Hohenheim University, Stuttgart ; Dipl.-Ing. J. Goldmann, DLG Test Center Technology, Groß-Umstadt Abstract Automatic guidance systems on tractors get more and more popular because they contribute to increase the efficiency of agricultural operations in several aspects. Aim of the investigation was to develop a sensing method to allow the assessment of automatic steering systems by quantifying cross track errors. A kinematic tachymeter (total station) was used to track a fixed point (prism) on the moving machine and function as a reference system. Tractors were tested in automatic mode on different routes. Cross track errors were calculated for describing the performance accuracy related to an ideal route. Due to its direct meaning the results for front and rear axle centers (interest points) are presented. The results show that for assuming a realistic positional accuracy of the GNSS system of around 2 cm that the errors are at least doubled in terms of lateral variability of driving trajectories. For higher speeds and partly more rough condition the errors are up to 4 fold the GNSS accuracy. The developed and presented measuring system was able to quantify the accuracy of the driving performance of tractors based on automatic steering systems. 1 Introduction Automatic guidance systems on tractors get more and more popular because they contribute to increase the efficiency of agricultural operations in several aspects [6, 11, 9, 13]. Systems with different accuracy levels are available for specific operational requirements. Applying fertilizers, driving in tramlines or implement control by tractors require different accurate driving abilities of the tractor. Automatic steering systems focussed first on tractor control but today they are able to precisely guide also implements passively or even actively [7, 11]. Furthermore today automatic steering systems are realised and offered through the ISO standard [5, 12]. Aim of the investigation was to develop a sensing method to allow the assessment of automatic steering systems by quantifying cross track errors. Today manufacturers often describe

2 their steering systems by stating the accuracy of the GNSS systems. However, the GNSS accuracy is important but how the tractor steering system processes this input data and how the tractor itself reacts in terms of driving behavior has to be questioned. The aim of the paper was to develop a sensing and testing method to compare various tractors with different steering systems under standardized conditions. 2 Materials and Methods An automatic steering system implemented in a tractor consists of several components: Global Navigation Satellite System (GNSS) with correction service, tilt sensor, navigation controller, terminal, interface to tractor, steering system type and design, tires, inflation pressure, tractor design, size and weight (kinematics and inertia forces) and the operation software. It is obvious that errors and malfunction or mismatching units will cause non accurate tractor driving. There are various potential sensing systems to determine the true driving path of a tractor. However, according to the ISO standard [2] the reference sensor to produce horizontal distance measurements required for the error calculations must be at least ten times more accurate than the assessment parameter of the vehicle being tested. One option is the use of landmarks as for example a steel cable or colored lines etc. [4]. But these aids are not easy to set up and to be utilised under realistic test conditions. Therefore we decided to use a kinematic tachymeter (total station) to track a fixed point (prism) on the moving machine (sampling 10 Hz). We used a Leica Viva TS15 imaging total station as a reference. The internal sensing consists of measuring two angles and the range to the prism in relation to the tachymeter location. The range is up to 3.5 km to a single prism and the static accuracy is 1 mm ppm at a resolution of 0.1 mm. The reflection prism was mounted in the tractor rear hitch providing free line of sight during the test. We calibrated the tachymeter by using fixed points with known absolute positions (UTM) in order to get absolute coordinates also as output data. The test course should consist of a repeatable route of travel comprised of one or more test course segments. The ISO standard suggests a realistic field surface [2]. Due to more easily achievable repeatability and comparability we decided to use the standard test course at the DLG Test Center with mainly concrete track surfaces. Furthermore, to simulate more dynamic driving conditions an existing rough track consisting of height alternating wooden bars were used. The DLG Test Program [8] describes the test conditions and suggests the following test routes. For straight A-B lines different routes were chosen (i) plain concrete (length 100 m),

3 (ii) with transversal offset (length 100 m), (iii) standardised height alternating wooden bars called rough track (length 100 m), (iv) transversally inclined track with constant angle and difference in track height 0.25 m (length 12 m) and (v) with tractor starting 90 degrees to test track orientation. Furthermore a contour track was tested as well as low GNSS positioning quality by losing the correction signal. The test conditions were documented by using date, time, ambient temperature, cloudiness, precipitation and quantified operation condition for the GNSS through the parameter DOPs (Dilution of Precision). Due to design variability of the rear hitch the position of the prism is machine dependent. In order to be able to transfer the absolute 3D prism coordinates to points of interest at the machine it was necessary to capture the dynamic 3D tilting of the machine during test ride. An inertial measurement unit (IMU) (VectorNav VN-100) was used to provide dynamic tilt information during testing (sampling 10 Hz). It was mounted beside the prism on the same metal frame. For the coordinate transformation of the prism and its tilting to the points of interest at the machine a simple geometric model was used. As points of interest two machine locations were chosen (i) center of the rear wheel axes and (ii) center of the front wheel axes. For the evaluation of the driving performance the cross-track errors of the points of interest were calculated from lateral deviations between the measured true path and the planned path. Fig. 1: Principle showing a total station (tachymeter) tracking the prism in the rear hitch of a test tractor (Photo Goldmann, DLG-Test Center) 3 Results and Discussion Due to the limited extent of this paper only some results can be presented. The direct deviation results from the prism are of low interest because the interpretation is difficult especially

4 when the prism location in relation to the machine is not standardised for all tractors. More powerful due to its direct meaning are the results from axle centers (interest points). The results are directly describing the longitudinal rear and front wheel tire trajectories by the lateral confidence interval (table 1). The results show that for a slow speed the confidence interval is ± 3 to 4 cm and for the higher speed around ± 8 cm. The results show that for assuming a realistic positional accuracy of the GNSS system of around 2 cm that the errors are at least doubled in terms of lateral variability of driving trajectories. For higher speeds and partly more rough condition the errors are up to 4 fold the GNSS accuracy. On the rough track with the slow speed we see the same lateral deviation for the rear wheels as for the plain surface but a higher deviation for the front wheels. It seems that the navigation system has more problems in keeping the machine on the straight track, because the front wheel movements are obviously of higher magnitude. It was expected to get this higher absolute and dynamic deviations at the front axis (active steering) compared to the rear axis. Table 1: Results from test runs showing the analysis of lateral deviations from ideal straight routes for a standard tractor of 7.76 t weight, 162 kw power and tires 540/65 R 30 and 650/65 R 42 with inflation pressure 1.2 bar Straight Line 8 km/h Straight Line 15 km/h Rough Track 8 km/h Point Prism Rear Front Prism Rear Front Prism Rear Front Repetitions Samples Mean deviation (m) Standard deviation (m) Maximum deviation (m) Confidence interval 95 % ±0.030 ±0.032 ±0.038 ±0.082 ±0.082 ±0.084 ±0.026 ±0.036 ±0.080 The confidence interval with its 95% range shows how accurate the tractor tires can be kept on or in one track. This is important for crop management e.g. spraying and fertilisation in tramlines which is done several times during the vegetation period.

5 The described investigations were used as a basis for discussing and setting up the final test program of the DLG Test Center for determining and assessing the performance of tractors using GNSS-based automatic steering systems [10, 8]. The assessment of the tractor behaviour is only a first step. Improved performance with passive or active implement control for several agricultural operations is expected. Further investigations must address the positional performance including implements. Implements often execute the relevant agricultural task while the tractor is providing draft force and other forms of power [7, 1, 3]. But Vellidis 2013 [13] could not confirm improvements by moving the positional interest points to the implement and hence documents potential for improvements. 4 Conclusions Accurate positional information from GNSS systems is no guarantee for accurate automatic steering of tractors, but a necessary requirement. How the positional data are processed and transformed into driving performance is depending on other active and passive system components as for example the controller algorithms as well as machine weight and tire types. The developed and presented measuring system was able to quantify the accuracy of the driving performance of tractors based on automatic steering systems. Acknowledgement The authors would like to thank T. Werner from Leica-Geosystems for many useful advices and the introduction to total stations. The project was conducted at the Max-Eyth Endowed Chair (Instrumentation & Test Engineering) at Hohenheim University which is partly granted by the Deutsche Landwirtschafts- Gesellschaft e.v. (DLG). References [1] Adamchuk V.I., Hoy R.M., Meyer G.E., Kocher M.F. (2007). GPS-based auto-guidance test program development. In: Proceedings European Conference on Precision Agriculture (ECPA) (editor Stafford J.V.) Skiathos, Greece, Wageningen Academic Publishers, Wageningen, The Netherlands, p [2] Anonymous (2012). ISO/FDIS Tractors and machinery for agriculture and forestry Test procedures for positioning and guidance systems in agriculture Part 2: Testing of satellite-based auto-guidance systems during straight and level travel

6 [3] Backman J., Oksanen T., Visala A. (2013). Applicability of the ISO network in a distributed combined guidance system for agricultural machines. Biosystems Engineering 114 p [4] Easterly D.R., Adamchuk V.I. (2008). Auto-Guidance Error Measurement Using a Visual Sensor. In: Proceedings ASABE Annual International Meeting Providence, Rhode Island, USA, ASABE, St. Joseph, MI, USA [5] Ehrl M., Auernhammer H. (2007). X-by-wire via ISOBUS communication network. Agricultural Engineering International: The CIGR Ejournal, Vol. IX, July 2007 [6] Ehrl M., Stempfhuber W., Demmel M., Kainz M., Auernhammer H. (2004). Autotrac- Accuracy of an RTK DGPS based autonomous vehicle guidance system under field conditions. In: Proceedings Automation Technology for Off-Road Equipment (ATOE), Kyoto, Japan, ASAE, St. Joseph, Michigan, USA [7] Gan-Mor S., Clark R.L., Upchurch B.L. (2007). Implement lateral position accuracy under RTK-GPS tractor guidance. Computers and Electronics in Agriculture 59(1-2) p [8] Goldmann J., Groth S. (2013). Prüfrahmen FOKUS-Test Automatische Lenksysteme (für landwirtschaftliche Maschinen). Unpublished, , DLG-Test Center for Technology and Farm Inputs, Groß-Umstadt, Germany [9] Griffin T. (2009). Whole-farm Benefits of GPS-enabled Navigation Technologies. In: Proceedings ASABE Annual International Meeting, Reno, Nevada, American Society of Agricultural and Biological Engineers (ASABE), 2950 Niles Road, St. Joseph, MI USA [10] Groth S. (2013). Entwicklung eines Messverfahrens zur Bewertung von automatischen Lenksystemen. MSc Thesis, Institut für Agrartechnik, Universität Hohenheim [11] Heraud J.A., Lange A.F. (2009). Agricultural Automatic Vehicle Guidance from Horses to GPS - How We Got Here, and Where We Are Going. In: Proceedings Agricultural Equipment Technology Conference, Louisville, Kentucky, USA, ASABE, St. Joseph, USA [12] Holtmann M. (2012). ISOBUS Auto Guidance - Mehr Flexibilität und universelle Einsatzmöglichkeiten. In: Proceedings LAND.TECHNIK, Karlsruhe, Germany, VDI-Verlag, Düsseldorf, Germany, VDI-Berichte 2173 [13] Vellidis G., Ortiz B., Beasley J., Hill R., Henry H., Brannen H. (2013). Using RTK-based GPS guidance for planting and inverting peanuts. In: Proceedings Precision Agriculture (ed Stafford, J. V.), Lerida, Spain, Wageningen Academic Publishers, Wageningen, The Netherlands, p