Autonomous robots in agricultural field trials

Transcription

1 Agricultural Field Trials Today and Tomorrow University of Hohenheim Stuttgart October 2007 Arno Ruckelshausen Faculty of Engineering and Computer Science COALA - Competence Center of Applied Agricultural Engineering 1

2 Overview 1. Introduction 2. Sensors in agriculture 3. Individual plant detection 4. Autonomous field robots 5. Conclusions 2

3 Overview 1. Introduction 2. Sensors in agriculture 3. Individual plant detection 4. Autonomous field robots 5. Conclusions 3

4 From the field to the single plant Conventional farming : Field-related figures Precision farming: Sub-field-related figures Single-plant farming : Plant-related figures 4

5 Weed in row cultures Source: Astrand

6 Technology challenge Precision Farming Individual Plant Farming GPS High precise GPS / landmarks Off-line sensors / Averaging on-line sensors Non-averaging on-line sensors Standard system technology Advanced system technology 6

7 Interdisciplinary approach for agricultural engineering Mechancis Application Knowledge Mechatronic System Electronics and Software Experience Experience Sensors Sensors Electronic systems Electronic systems Software Software / Algorithms 7

8 Overview 1. Introduction 2. Sensors in agriculture 3. Individual plant detection 4. Autonomous field robots 5. Conclusions 8

9 Key issues for sensors in agriculture Robust Smart Low-cost Interpretation of sensor signals Sensor Fusion Robust algorithms Behaviour guidelines 9

10 Example 1: Optoelectronic distance measurement Applications: Measurement of crop height and density, stalk detection Triangulation sensor Optoelektronisches Sensorsystem zur Messung der Pflanzenbestandsdichte G. Thösink, J. Preckwinkel, A.Linz, A. Ruckelshausen, J.Marquering; Landtechnik 59, S.78-79,

11 Results Velocities up to 10 km/h Spotsize in the mm 2 -range Statistical online analysis Sensor signals Reduced data 11

12 Example 2: Imaging sensor system ( 1-bit-imaging ) Test setup (laboratory) Maize plants (field) direction of motion side view trigger latitude 12

13 13

14 Example 3: Sensor for automatic cutting length variation Online-measurement of the degree of maturity for maize plants 14

15 Method: Optoelectronic NIR sensor system Pulsed LEDs (spectral characteristics) Microcontroller-based solution (embedded system) Self-cleaning effect 15

16 Field Tests 16

17 Wavelength Agricultural Field Trials Today and Tomorrow / October 2007 Example 4: Hyperspectral Imaging Position Example: 4 LEDs with different wavelengths 17

18 System integration: spectral imaging with ImSpector/CMOS-camera "Weed detection based on spectral imaging systems with CMOS cameras ; S. In der Stroth, B.Ramler, A.Linz, A.Ruckelshausen ; 4 th European Conference on Precision Agriculture ECPA, Berlin, Programme book, ISBN , Juni

19 Application: Harvesting, quality control Potatos, stones, earth, plants Pointwise spectral analysis and online image processing Förderung Arbeitsgruppe Innoavite Projekte beim Miniserium für Wissenschaft und Kultur, NIedersachsen 19

20 Sensor Fusion: Properties of plants (examples) Reflection [%] 80 Spectral properties (reflection of various green plants) Wavelength [nm] top view Geometrical properties side view Maize Forgetmenot Combination of different sensors with varying selectivties 20

21 Architecture of a sensor fusion based mechatronic system Height profile sensor Distance sensor Sensor Position Sensing μc μc μc Host μc CAN - Bus Interface μc μc μc Hoe Sprayer Actuator 21

22 Agricultural Field Trials Today and Tomorrow / October 2007 First application of sensor fusion in weed control (1998) Application: Mechanical intra-row weed control Multi-sensor system / microcontroller-based architecture Transversal cycloid hoe ( Querhacke ) Plant database " Microcontroller-based multi-sensor system for online crop/weed detection, A.Ruckelshausen, T.Dzinaj, F.Gelze, S.Kleine Hörstkamp, A.Linz, J.Marquering, Proceedings of the International Brighton Conference "Weeds", 1999, pages

23 Overview 1. Introduction 2. Sensors in agriculture 3. Individual plant detection 4. Autonomous field robots 5. Conclusions 23

24 Technologies for individual plant detection System technology (real-time, embedded systems, CAN-bus) Software (examples: data acquisition, algorithms, testing, communication) User interface Sensors for crop detection (examples: photo diode arrays, spectral imaging) Positioning with GPS and other sensors (encoder) Power management Alarm unit (example: voltages, temperature, dust) Mechanical mobile unit Test environment 24

25 System architecture for individual plant detection (sensors) Height Profile Sensor Distance Sensor Sensor User Interface μc μc μc Host μc CAN - Bus μc μc GPS Spectral Imaging 25

26 GPS technologies Application in maize fields: typical distance of 2 maize plants is 8 to 12 cm GPS accuracy better than 5 cm (RTK/DGPS) Typical measured RTK/DGPS accuracy: ± 2 cm Interpolation of two GPS signals with encoder information GIS-tool OpenJUMP (open-source application) RTK/DGPS system MS750 from Trimble 26

27 Mobile sensor unit in a maize field 27

28 Visualization of data with the GIS-tool OpenJUMP 28

29 Results 4 different runs in the same maize row Vertical variations due to different tracks of the mobile unit Identification of an individual plant Reidentification of an individual plant 29

30 Hey Joe, you look better today than last week. Moreover you have grown more than 2 cm. See you next week! Joe Joe 30

31 Overview 1. Introduction 2. Sensors in agriculture 3. Individual plant detection 4. Autonomous field robots 5. Conclusions 31

32 Autonomous Service Robots: Field Robots Cost reduction, environmental protection, jobs Technologies: Sensors, software, algorithms, actuators, system technology, vehicles, safety, application related aspects, user interface Fun, as for example: International Field Robot Event 32

33 International Field Robot Event Initialized by Wageningen University. Interdisciplinary teams from all over the world compete. Tasks: Navigation through straight and curved maize rows Make a turn at the end of the row into the next or another row Counting of the plants Aufgabe 1 Aufgabe 2 Weed control 75 cm 75 cm Freestyle 33

34 Sensor fusion concept Sensors AVRcam IR Sensor Flex Sensor Ultra Sonic Sensor Gyroscope Hall Sensor Photodiodes Light sensor Navigation Counting yellow balls Robot tasks Drawing white line Speed race Finding holes Watering flowers Turn on end of row 34

35 Architecture of an autonomous field robot (Amaizeing) 35

36 Smart low-cost camera CMUCam Integrated microcontroller for image processing Options for reduced data (example: windowing) Tracking options (example: color tracking) Resolution up to 160 x 255 Pixel Low cost solution (ca. 140 ) 36

37 Navigation: Examples Navigation within the row Turnaround 37

38 Integration of actuators Drawing a white line Watering flowers 38

39 User Interface System tests Test of sensors WLAN Strategies (Software) Algorithms Parameter Documentation GUI Field Robot Amaizeing 39

40 Communication Learn-Mode, transfer of software and parameters, electronic documentation, safety Platform for teleservice, robot swarms 40

41 From experimental studies to prototypes: first steps Autonomous tractor: University of Kopenhagen (Cooperation Querhacke ; Osnabrück) Autonomous vehicles for weed control (examples): HortiBot (Aarhus, DK) and Weedy (Osnabrück, G) 41

42 A concept for an autonomous robot for field trials ( BoniRob ) Key technologies: Application orientation, sensor systems, robust software, safety Complete analysis: field, site, plant Analysis of individual plant parameters Source: Satconsystem Kinderfinder ; 42

43 Overview 1. Introduction 2. Sensors in agriculture 3. Individual plant detection 4. Autonomous field robots 5. Conclusions 43

44 Conclusions Autonomous robots for field trials: Interdisciplinary challenge Key technologies: Robust software, sensor technologies, safety Individual plant detection is possible (first tests have been performed) First experiences with autonomous field robots are available Major problem of autonomous field robots: Robust navigation Next steps: Development of a robust platform for agricultural field trials ( BoniRob ) Experimental period : tests, improvements, extensions Integration of actuators 44

45 Autonomous Field Robots Agricultural Field Trials Today and Tomorrow / October 2007 Hall 16 Stand 16D17 / DLG Hall 14 Stand A16 / Amazonen-Werke Field Robot Event2008 June Osnabrück/Germany 45