GNSS-Based Auto-Guidance Test Program Development

Transcription

1 ECPA (Skiathus( Skiathus,, Greece) June, GNSS-Based Auto-Guidance Test Program Development Viacheslav I. Adamchuk George E. Meyer Roger M. Hoy Michael F. Kocher George E. Meyer Michael F. Biological Systems Engineering University of Nebraska-Lincoln Background Auto-guidance (auto-steering) systems provide numerous benefits Performance of different systems depends on internal and external factors Quality of auto-guidance performance must be quantified in repeatable manner ASABE has two active projects: GPS Dynamic Test Standard (X) Auto-Guidance Test Standard (X) Outline Why do auto-guidance systems perform differently? What is guidance error? auto-guidance field day demo review pilot test of different systems using Nebraska Tractor Test Laboratory s test track instrument development and future plans Agricultural Vehicle Guidance Reference method Local triangulation Crop-based methods Mechanical feelers Laser row tracking Machine vision GNSS-based guidance Level of assistance Navigation aids Lightbar parallel tracking Auto-guidance Auto-steering Autonomous vehicles Field robots GNSS-Based Auto-Guidance Popular GPS Solutions Positioning Sensor DGPS (< in) RTK (< in) Terrain Compensation Gyroscope Multiple Units Vehicle Control Mechanical Hydroelectric Single Frequency Receivers WAAS, EGNOS, Beacon OmniSTAR VBS, John Deere SF Subscription/Free (sub-meter accuracy) Dual Frequency Receivers OmniSTAR HP/XP, John Deere SF Subscription (decimeter accuracy) RTK Receiver Base Station (centimeter accuracy)

2 Auto-Guidance Applicability What does ± actually mean? Sub-meter accuracy - ft year-to-year ft pass-to-pass Mechanical steering Decimeter accuracy in year-to-year in pass-to-pass Electro-hydraulic controls Centimeter accuracy in year-to-year in pass-to-pass Base station Tillage/Disking Spraying/Spreading Harvesting Seeding Mapping Planting Cultivating Bedding Strip Tilling Drip Tape Placement Land Leveling Topographic Mapping Cross-track error Nature of the test Static vs. dynamic Duration of the test Pass-to-pass vs. long-term Definition of the test Precision versus accuracy Statistic used % (σ) vs. 9% (σ) prediction Type of error Positioning vs. guidance error Auto-Guidance Error Auto-Guidance Field Day GPS Positioning Error Vehicle Dynamics Guidance Error Implement Tracking Field Conditions August, ARDC, Mead, Nebraska Field Demonstration Demonstration Course Pull-type cart J-type course Terrace crossing Coulter marker RTK-level GPS position logging Contour pass Straight pass Starting point

3 Straight Pass Contour Pass 9 9 Ea sting, m Ea sting, m 9 L A A A Error Definition (X,Y) α LA ε (XA,YA) L LA L = L + L L L cosα cosα = sinα = cos α ε = L A sinα (X,Y) A ε = L + L L L L A A A L + L L L A AL A Cross-Track Error, cm Error Distributions Exact backtracking Curved Path Straight and Level Path Straight and Sloped Path System System System System System Cross-track Error (cm) Error Distributions Signed error distribution Signed Error Unsigned Error Normal Model Rayleigh Model Zero Error Average Error Median Absolute Error Mean Absolute Error RMSE Unsigned error distribution Cummulutive Probability Cumulative Distribution Signed Error Unsigned Error Normal Model Rayleigh Model Zero Error 9% Threshold Cross-Track Error, cm

4 Producer-Viewed Differences NTTL Track Testing Interface and ease of use Modes of operation Installation time and options Setup and calibration procedure Cost and possible upgrades Versatility and secondary use Technical support Reference GPS antennae Two straight parallel passes RTK Base Station Linear potentiometer Trigger Test cart Linear Potentiometer System (LPS) Pilot Test R Nuts Vertical support Resistor array washers Common array washers Rw Rw Rw Rw Rw Rw V Steel rod PVC tubing PVC spacer System A Centimeter-level accuracy System B Decimeter-level accuracy LPS based analysis GPS based analysis Straight passes Three -min runs Two consecutive days Urban environment Concrete pavement Trigger.9 cm Test Sequence North 9 triggers South triggers Spacing. m Swath. m Test Run Test Run Test Run Test Run A-B line

5 LPS Output Number of occurances North North North South South South System A Relative position, cm Number of occurances North North North South South South System B Relative position, cm Northing (North pass), m Reference GPS Output North South Linear (South) Linear (North) R =.9 R = Northing (South pass), m Error Definition Standard error Difference between each trigger hit position and an average (balanced in terms of direction of travel) of all relevant hit positions calculated for each trigger separately (within min for short-term error and all test data for long-term error) Standard error of a simple linear regression for the GPS data Cross-track error Expected difference between two tracks randomly selected from either a short-term or long-term pools of data Theoretically, cross-track error equals to standard error multiplied by square root of System A centimeter accuracy System B decimeter accuracy System A - GPS reference (short-term error) System A - LPS reference (short-term error) System B - GPS reference (short-term error) System B - LPS reference (short-term error) System A - GPS reference (long-term error) System A - LPS reference (long-term error) System B - GPS reference (long-term error) System B - LPS reference (long-term error)

6 Short-term term Error (Pass-to to-pass) Long-term Error (Day-to to-day) % 9% % % % % % % % % 9% probability Cumulative short-term error distribution System A (LPS) System A (GPS) System B (LPS) System B (GPS) % 9 Error, cm % 9% % % % % % % % % 9% probability Cumulative long-term error distribution System A (LPS) System A (GPS) System B (LPS) System B (GPS) % 9 Error, cm Standard Related Discussion Items Optical Sensor Development Test location Surface conditions Clarity of sky Test course segments Method of measurement Reference receiver Total station Mechanical (contact) sensor Optical (non-contact) sensor Test sequence PDOP requirements Pass-to-pass test Day-to-day test Error terms Dealing with bias Parametric and nonparametric estimates Machinery selection Tractors Sprayers Video camera with internal processor Direct hard mounting to the tractor s chassis Locating the center of a painted line Sensor to within calibration mm Test Plans Measurement close to the drawbar pivoting point Three -min test run sequences repeated three times Actual cross-track error calculation Non-parametric statistics only Three test segments Linear flat Curved flat Curved sloped Line acquisition test trial E:mail: vadamchuk@unl.edu