Precision Guidance of Agricultural Tractors for Autonomous Farming

1
Precision Guidance of Agricultural Tractors for
Autonomous Farming

• R. Eaton1
• J. Katupitiya2
• K. W. Siew2
• K. S. Dang2
• 1School of Electrical Engineering and
  Telecommunications
• 2School of Mechanical and manufacturing
  Engineering
• The University of New South Wales
• Sydney, Australia

2
Contents

• The problem
• Automation in farming a background
• Precision Agriculture
• Autonomous Machinery Operations
• Tractor-Implement Guidance
• The Farming System
• Our toys Autonomous Farming Machinery
• Looking to the future

3
The Problem

• A decrease in the size of the skilled and
  unskilled labour workforce.
• In Australia, numbers working in agricultural
  industry decreased from 330,000 by 100,000 since
  2002 farm hands and heavy farm equipment
  operators some of hardest hit! National Farmers
  Federation
• Change in farming culture has lead to an increase
  in mechanisation chicken or egg problem?
• Increased emphasis on corporate farming
• Decrease in traditional family farming
• Large/vast crop areas
• More competition and on a global scale
• Need for farming efficiency and productivity to
  survive.
• These factors lend themselves to an increase in
  the utilisation of automation.
• Seen more as a necessity to compete/survive.

4
The Problem

• How to put automation in place for large-scale
  farming solutions? (broad-acre crops)
• Safety
• Efficiency
• Productivity
• Integration into farming landscape
• Gradual deployment necessary and desired
• Uptake of automation?

5
Contents

• The problem
• Automation in farming a background
• Precision Agriculture
• Autonomous Machinery Operations
• Tractor-Implement Guidance
• The Farming System
• Our toys Autonomous Farming Machinery
• Looking to the future

6
Automation in farming

• Simple automation can be traced back to
  pre-1800s, e.g. the cotton gin in 1793.
• A fairly steady stream of inventions since then,
  but not true robots, none could think and act
  for themselves.
• Only in the quite recent past that efforts have
  been made to develop agricultural robots.

7
Precision Agriculture

• Precision Agriculture (PA) is about doing the
  right thing, at the right place, in the right
  way, and at the right time (Wikipedia).
• It refers to the use of methods and resources to
  address in-field variability of factors which
  affect crop growth, such as soil type, soil
  nutrient levels, type and levels of fertiliser,
  weed growth, etc.
• PA implies the use of new technologies such as
  GPS, and appropriate sensors.
• A true Precision Autonomous Farming system must
  be achieved with a mix of Precision Agriculture
  and precision autonomous machinery.

8
Autonomous Machinery Operations

• Most research has been focused on tractor
  guidance, leading to commercialisation.
• Some additional work has been done for other
  broad-acre farm operations, including harvesting,
  spraying, and weeding in their infancy stages.

9
Tractor Guidance

• GPS-based guidance systems commonly available in
  commercial tractors, with sub-inch accuracy.
• Increased accuracy from sensing accuracy.

• Laser guided tractors (useful for obstacle
  avoidance)

10
Tractor Guidance Problems

• Robustness - control based on the assumption of
  no-slip.
• Auto-steer only - without propulsion control.
• BIG PROBLEM - does not take into account any
  attached implements!
• It is the attached implement that is typically
  performing the agricultural task precision
  guidance is required more so of the implement!

11
Tractor-Implement Guidance

• Research into the robust guidance of tractors
  under slip conditions is infant no implement
  assumed as yet.
• Fang et. al. have been quite active in this area
• Research into the guidance of tractor-implement
  systems has more of a past no-slip condition is
  assumed however (not realistic for agricultural
  environments)
• Much work on tractor-trailer control, e.g. Wang
  et. al. and Hingwe et. al.
• Agrawal et. al. have looked at a steerable
  trailer.
• It is the aim of our research to design, build
  and control an active implement to robustly
  achieve the desired precision.

12
Contents

• The problem
• Automation in farming a background
• Precision Agriculture
• Autonomous Machinery Operations
• Tractor-Implement Guidance
• Crop weeding
• The Farming System
• Our toys Autonomous Farming Machinery
• Looking to the future

13
The Farming System

• Outputs
• crop yield and quality, efficiency of crop
  operations
• Inputs
• type/amount of seed/fertilise/pesticides, fuel,
  land geometry, available resources

14
The Farming SystemA. Farming Layout System

• Aim to produce good structure and optimal traffic
  conditions for the machinery.
• Determined via the land geometry, contour maps,
  crop type, available resources, soil type and
  condition. This information is drawn from the
  Precision Agriculture Data Set (PADS).
• Determining optimal crop layout will produce a
  Precision Farming Data Set (PFDS).

15
The Farming SystemB. PFDS and PADS

• Precision Farming Data Set (PFDS)
• Describes the navigation and spatial accuracy
  requirements.
• The basis for other farming machinery sub-systems
  where spatial accuracy is required (route map in
  broad acre farming).
• Current farming trends require a 2cm accuracy in
  the lateral direction of crop rows.
• Precision Agriculture Data Set (PADS)
• Works in collaboration with the PFDS to ensure
  the agronomy requirements of the crop are
  satisfied.
• Evolving data set which develops with crop
  growth, when crop sensing and follow-up
  operations take place.
• E.g. specifies fertiliser type, application
  rates, crop growth, and soil conditions, tied to
  spatial data.

16
The Farming SystemC. Automated Machinery
Operations

• Comprised of smaller sub-systems each dealing
  with more specific agricultural tasks.
• Crop seeding
• Crop sensing
• Follow-up operations
• Harvesting

17
The Farming SystemC. Automated Machinery
Operations

• Crop seeding
• Arguably where most spatial accuracy is required.
• Seeding takes place via an attached seeding
  implement.
• Significant disturbance forces can effect ability
  to maintain accuracy ground engagement and
  gravitational forces.
• Crop sensing
• Crop growth, soil moisture, weed
  prevalence/growth, etc.
• Information fed into the PADS so that it enhances
  the efficiency and accuracy of follow-up
  operations.
• Can be done via machinery using the PFDS, or
  perhaps via aerial means.

18
The Farming SystemC. Automated Machinery
Operations

• Follow-up operations
• Application of fertiliser, pesticides, herbicides
  during growth.
• Such operations make use of both the PFDS for
  spatial accuracy and PADS for agronomy data such
  as fertiliser rates, and weed treatment dosages.
• Required accuracy is not as great in general.
• Harvesting
• Autonomous and coordinated harvesting and grain
  collection machinery can traverse the crop via
  the use of the PFDS.
• Crop yield and quality measurement done
  on-the-fly and provided to the PADS providing
  spatially sorted information.

19
The Farming SystemD. Farming Software System
20
Contents

• The problem
• Automation in farming a background
• Precision Agriculture
• Autonomous Machinery Operations
• Tractor-Implement Guidance
• Crop weeding
• The Farming System
• Our toys Autonomous Farming Machinery
• Looking to the future

21
Our Toys Autonomous Farming MachineryA.
Precision Autonomous Seeding

• Research is focused on the precise guidance of an
  active (not passive) seeding implement pulled by
  an autonomous tractor.
• Progress
• Design of an active (steerable and with
  propulsion) seeding implement. In construction
  stages now.
• Continued research in the precise and robust
  guidance of an agricultural tractor. Important
  work includes modelling and simulation of
  trajectory tracking controllers.

22
Our Toys Autonomous Farming MachineryA.
Precision Autonomous Seeding
23
A.1 Autonomous Tractor Test-bed

• Sensors for navigation give accurate position and
  orientation information - x, y, z, and roll,
  pitch, yaw.
• Dual differential RTK GPS (2cm and 20cm)
• IMU
• Tilt sensor
• Differential GPS obtained via the use of a third
  base station receiver.
• IMU mounted precisely, and gives short term
  position tracking in between GPS measurements and
  as a back-up to the GPS.

24
A.1 Autonomous Tractor Test-bedNavigation
systems testing under manual control
25
A.1 Autonomous Tractor Test-bed

• Additional sensors for low-level sub-system
  control
• Wheel encoders on rear wheel used to measure (and
  thus control) wheel speed, and also used in
  collaboration with the navigation sensors to
  provide information about wheel slip.
• Linear potentiometer used to measure (and thus
  control) front wheel steering angle.

26
A.1 Autonomous Tractor Test-bedOther Features

• Safety
• Watchdog system used to halt all mechanical
  operations of the tractor under all fault
  conditions.
• Remote start-up circuit
• Real-time software system
• Remote and on-board computer communicate via
  wireless internet.
• On-board computer driven by RTLinux, with time
  critical sensing and control tasks run in
  real-time threads.

27
A.1 Autonomous Tractor Test-bedControl Systems

• Low-level control in place to ensure propulsion
  and front wheel steering angle control.
• A high level path/trajectory tracking controller
  has to be implemented to provide the low-level
  controllers with desired speed and steering
  angles.
• Successful simulation of a robust (assuming slip)
  nonlinear controller.
• The controller is ready to be implemented and
  tested on the existing John Deere tractor.

28
A.1 Autonomous Tractor Test-bedManual Control
29
A.1 Autonomous Tractor Test-bedTrajectory
Tracking Control Slip velocities included
30
A.1 Autonomous Tractor Test-bedModelling of a
tractor-trailer-implement

• Modelling work has been extended to include an
  additional trailing element such as that used
  when carrying seed and fertiliser.
• Features of the model
• Complete dynamic model of the system under the
  influence of realistic disturbances.
• Highly nonlinear
• Inputs tractor and implement steering, and
  propulsion of the tractor
• Output position and orientation of the
  implement used for seeding.

31
A.1 Autonomous Tractor Test-bedModelling of a
tractor-trailer-implement
32
Our Toys Autonomous Farming MachineryB.
Non-Herbicidal Weeding Machine

• Weed eradication/minimisation takes place 2-3
  weeks into crop growth.
• Two stages
• Weed detection fairly crude at the moment but
  work continues to discriminate between different
  type of weeds.
• Weed destruction mostly done by herbicide
  spraying not optimised for different weed
  types.
• The use of non-herbicidal means of destruction is
  a much sought after solution.

33
Our Toys Autonomous Farming MachineryB.
Non-Herbicidal Weeding Machine

• Non-herbicidal (electrocution) based destruction.
• PFDS/Laser/Vision/GPS guided navigation.
• Small footprint.

34
Contents

• The problem
• Automation in farming a background
• Precision Agriculture
• Autonomous Machinery Operations
• Tractor-Implement Guidance
• Crop weeding
• The Farming System
• Our toys Autonomous Farming Machinery
• Looking to the future

35
Looking to the future

• Immediate
• Robust, precise and autonomous guidance of the
  tractor plus active seeding implement.
• Investigation of efficiency/performance of the
  non-herbicidal weeder possibly different forms
  of weed destruction.
• Long term
• Teams of autonomous and coordinated farm
  vehicles, tirelessly, efficiently, and safely
  performing the farming tasks.