66 GPS Accuracy: Enhancement Decisions
66.1 Start With the Situation
A basic receiver cannot meet the required accuracy on its own. The team must decide whether corrections, RTK, or a different product promise is justified by the site, safety need, and operating cost.
66.2 Overview
This route compares enhancement methods and carries the choice into cost, safety, and deployment evidence.
This is part 2 of 2. Review GPS Accuracy: Error Budgets when you need the first route.
66.3 Learning Objectives
By the end of this chapter, you will be able to:
- explain differential GPS corrections
- compare DGPS and RTK deployment trade-offs
- justify accuracy technology from cost and safety evidence
66.4 Chapter Roadmap
Follow the original sections below in order. They begin at the reviewed split boundary and keep every worked example, figure, check, and supporting banner with the section that owns it.
66.5 Differential GPS (DGPS)
The first enhancement tier is differential correction: use a known reference to cancel shared outdoor GNSS errors. Read Figure 66.1 from the common satellite observations to the surveyed base, then follow the calculated correction over the data link to the rover. The useful question is which errors are sufficiently correlated across that baseline to be cancelled.
Figure 66.1 shows the base using its true coordinates to compare against its GNSS solution and distribute a local correction that the rover applies. Satellite clock, orbit, and broad atmospheric effects can be shared; rover-local multipath and receiver noise are not magically removed. The system therefore improves the error budget while still requiring correction age, link state, and local quality to remain visible.
DGPS Principle:
-
Base Stations at Known Locations:
- Precisely surveyed reference positions
- Distributed around the world
-
Base Stations Calculate Local Errors:
- Position themselves using GPS
- Compare to known true position
- Model local GNSS errors
-
Transmit Corrections:
- Broadcast corrections out-of-band (radio, internet)
- Local receivers apply same corrections
- Accuracy: 1-3 meters (vs 5-10m for standard GPS)
Historical Note:
- Came about due to US “Selective Availability” (SA)
- SA: Intentional GPS errors for security (civilians got degraded service)
- SA turned off in May 2000
- DGPS still valuable for removing natural errors
The historical change in Figure 66.2 separates intentional degradation from the natural errors that remain. Read left to right across the 4:00 AM marker: the wide, noisy band before the change collapses quickly, then settles around a smaller but non-zero error floor. That residual is why differential and later carrier-phase techniques still have a role.
Figure 66.2 shows a stylized reconstruction, so use it to understand the step change rather than to infer a particular receiver trace. The remaining band represents the continuing accuracy problem: atmosphere, orbit knowledge, clock residuals, geometry, multipath, and receiver noise. This closes the DGPS section by connecting correction architecture to an honest product state rather than a promise of perfect coordinates.
66.6 Advanced GPS Techniques
The three enhancements solve different positioning errors, so accuracy alone is not enough to select one. The comparison diagram in Figure 66.3 separates faster acquisition, centimetre precision, and ionospheric correction before the chapter applies those choices to a field system.
The A-GPS panel in Figure 66.3 starts with an ephemeris download of about 20 seconds per satellite and uses cellular or Wi-Fi assistance to improve Time To First Fix, while remaining in the 3–5 m class. RTK (Carrier Phase) instead combines a Base station + rover pair and continuous satellite lock to reach 1–5 cm. Multi-Frequency compares L1, L2, and L5 so ionospheric delay can be removed, producing metre-level improvement without making the RTK infrastructure claim. This comparison keeps speed, precision, and correction mechanism distinct.
1. Assisted-GPS (A-GPS):
Problem: Downloading ephemeris (satellite orbital data) takes ~20s per satellite. Solution: Get ephemeris via cellular or Wi-Fi data connection. Benefit: Faster Time To First Fix (TTFF), works indoors better.
2. Carrier Phase Positioning (RTK):
GPS signal wavelength: ~19 cm (L1 band at 1575 MHz). Measure phase of carrier wave (not just code). Accuracy: Centimeter-level (2-5 cm). Requirements:. Very accurate initial location. Observe satellites for extended period. Maintain continuous lock (no “cycle slips”). Applications: Surveying, precision agriculture, autonomous vehicles.
3. Signal Structure:
Multiple frequency bands: L1 (1575 MHz), L2 (1227 MHz), L5 (1176 MHz). Data rate: Only 50 bits per second (very slow!). Uses spread-spectrum (CDSS) for signal strength and interference resistance.
66.7 RTK GPS in Precision Agriculture
RTK is the chapter’s clearest example of accuracy as a business and safety requirement rather than a luxury specification.
Understanding when augmentation (DGPS, RTK) is needed is critical for IoT system design. Let’s examine a real-world precision agriculture case where centimeter-level accuracy is mandatory.
Problem: Modern farms use autonomous tractors for planting and harvesting. Crop rows are spaced 30-76 cm (12-30 inches) apart. Standard GPS (5-10 meter accuracy) would cause the tractor to:
Destroy crops: Driving over planted rows instead of between them. Miss coverage: Leaving gaps in planting or harvesting. Overlap excessively: Wasting seeds, fertilizer, and fuel.
Why Standard GPS Fails:
| Technology | Accuracy | Can Navigate 50cm Rows? | Result |
|---|---|---|---|
| Smartphone GPS | 5-10 m | No | 10× too imprecise—tractor wanders across 20 rows |
| DGPS | 0.5-1 m | No | Still 2× too large—damages crops on both sides |
| RTK GPS | 1-2 cm | Yes | Pass-to-pass accuracy maintains lane position |
Solution: RTK GPS provides sub-2cm accuracy enabling:
Straight-line guidance: Maintain row centerline within ±2 cm over 1 km runs. Repeat passes: Return to exact same tracks on subsequent operations (planting → spraying → harvesting). Nighttime operation: Autonomous 24/7 operation without visual references. Reduced overlap: Typical 5-10% overlap reduced to 1-2% (saves $25,000-$50,000/year on 1000-acre farm in fuel, seeds, chemicals).
The crop-row requirement now needs an end-to-end correction path, not just an RTK accuracy number. The system diagram in Figure 66.4 shows the infrastructure that must stay intact for an autonomous tractor to hold its line.
In the system diagram Figure 66.4, the Base Station occupies a surveyed known location and observes the same GPS Sat 1, GPS Sat 2, and GPS Sat 3 as the tractor. It computes error corrections and sends Carrier Phase Corrections over the 900 MHz Radio Link. The Rover Receiver on the autonomous tractor applies them to produce a 1–2 cm fix between the marked 50 cm rows. That dependency chain explains why loss of satellite lock, base survey error, or a blocked correction link must trigger the fallback behavior discussed next.
Decision Framework for IoT Location Systems:
| Application Requirement | Recommended Technology | Typical Cost | Justification |
|---|---|---|---|
| Row-crop farming (30-76 cm rows) | RTK GPS | $15,000-$25,000 | 1-2 cm accuracy required; ROI in 1-2 years from reduced overlap |
| Broadcast fertilizing (no rows) | DGPS | $2,000-$5,000 | 0.5-1 m sufficient to avoid coverage gaps and excessive overlap |
| Livestock tracking (which pasture?) | Standard GPS | $50-$200 | 5-10 m accuracy adequate; just need geofence boundaries |
| Construction grading (±2 cm elevation) | RTK GPS + tilt sensors | $20,000-$40,000 | Vertical accuracy critical for drainage and leveling |
| Drone surveying (topographic maps) | RTK GPS or PPP | $5,000-$15,000 | Centimeter-level required for accurate 3D models |
| Fleet delivery tracking | Smartphone GPS | Free (built-in) | Know which city block vehicle is on; 5-10m sufficient |
Key Insight: Match accuracy to application—don’t over-engineer! RTK GPS costs 10-50× more than standard GPS. Only use it when centimeter-level precision delivers measurable ROI.
Common Pitfalls When Designing RTK Systems:
Requires Continuous Radio Link (Base ↔ Rover):. Problem: Hills, trees, buildings block 900 MHz/2.4 GHz radio. Range limit: Typically 10 km line-of-sight, much less in rugged terrain. Solution: Deploy multiple base stations or use NTRIP (internet-based corrections).
Loses Accuracy if Satellite Lock Breaks:. Problem: Driving under trees, bridges, or in narrow canyons causes “cycle slips”. Recovery time: 30 seconds to several minutes to re-establish centimeter fix. Fallback: System degrades to DGPS (1m) or standard GPS (5-10m) during outages.
Expensive and Power-Hungry:. Base station: 5-15 watts continuous power (requires AC or large solar panel). Rover receiver: 2-5 watts (vs. 0.5W for standard GPS). Total system cost: $15,000-$25,000 (base + rover + radio + installation).
Requires Expertise to Set Up:. Base station must be surveyed to ±1 cm accuracy (professional surveyor). Radio frequency coordination (avoid interference with other farm equipment). Antenna placement (clear sky view, stable mounting, lightning protection).
When RTK is NOT Worth It:
Wrong: Use RTK to track shipping containers (5-10m GPS sufficient to know “which warehouse”). Wrong: Track hospital wheelchairs indoors (GPS doesn’t work indoors; use BLE beacons instead). Right: Autonomous tractors planting 50cm rows ($40,000/year savings justifies $20,000 equipment).
Checkpoint: Enhancement Selection
You now know:
- DGPS improves nearby outdoor receivers by correcting shared errors.
- RTK needs continuous carrier-phase lock, correction delivery, and clear enough sky to keep centimeter confidence.
- The right selection record states the accuracy tier, cost, power, fallback mode, and failure message.
66.8 Knowledge Check
Illustrative Cost Comparison:
Two modeled IoT deployments illustrate the importance of matching accuracy to application. Treat these numbers as planning assumptions for comparing orders of magnitude, not as claims from a named deployment:
| Scenario | Requirement | Technology | Cost | ROI |
|---|---|---|---|---|
| Last-Mile Delivery Fleet (200 trucks) | 10m accuracy to know which city block | Standard smartphone GPS | $19,200/year (cellular data only) | $45,000/year fuel savings from optimized routing |
| Autonomous Tractor Farm (4 tractors, 2,000 acres) | 2cm accuracy for 50cm row spacing | RTK GPS system | $40,800 upfront + $2,400/year | $117,000/year savings (8% overlap reduction + 24/7 operation), 4.3-month payback |
Key insight: In this model, the delivery fleet would waste money on unnecessary RTK precision. The tractor deployment would carry large operational risk if standard GPS were used for row guidance.
GPS Technology Selection by Application:
| Application | Accuracy Needed | Recommended Technology | Typical Cost/Unit | Key Constraint |
|---|---|---|---|---|
| Shipping container tracking | 10-50 m (city-level) | Standard GPS | $50-150 | Battery life (years), not accuracy |
| Delivery fleet management | 5-10 m (street-level) | Standard GPS + SBAS | $0-200 | Free GPS in phones sufficient |
| Autonomous lawnmower | 10-20 cm (avoid flower beds) | RTK GPS or PPP | $1,500-4,000 | Needs obstacle avoidance sensors too |
| Precision seeding tractor | 1-2 cm (50cm row spacing) | RTK GPS with base station | $8,000-15,000 | Must maintain continuous phase lock |
| Construction grading | 2-5 cm (drainage requirements) | RTK GPS + tilt sensors | $15,000-30,000 | Vertical accuracy critical |
| Drone surveying | 5-10 cm (topographic mapping) | RTK or PPP | $3,000-8,000 | Weight/power constraints |
| Personal fitness tracker | 3-10 m (route tracking) | Standard GPS (smartphone) | Free | Battery life, not accuracy |
| Geofencing (entry/exit zones) | 10-30 m (zone boundaries) | Standard GPS | Free | Zone design compensates for accuracy |
Cost-benefit decision tree:
Is <5m accuracy ESSENTIAL for function?. No → Use standard GPS ($0-200/unit). Yes → Continue to question 2.
Is <10cm accuracy ESSENTIAL?. No → Use DGPS or SBAS ($0-500/unit for receiver). Yes → Continue to question 3.
Can you deploy a base station?. Yes → Use RTK GPS ($8,000-15,000/unit + $10,000 base shared across all units). No → Use PPP ($2,000-5,000/unit + $50-200/month subscription).
Is real-time required?. Yes → Must use RTK or real-time PPP (requires radio or cell connection). No → Can use post-processed PPP (acceptable for surveying, not for autonomous vehicles).
Power/cost tradeoff:
Standard GPS: 0.5W continuous. RTK GPS: 2-5W continuous (processing carrier phase). For battery-powered devices, accuracy comes at 4-10× power cost.
Calculate the return on investment for RTK GPS in precision agriculture:
Insights: Under the calculator’s default assumptions, payback falls in the 4-8 month range because overlap reduction and labor assumptions dominate the result. Smaller farms, lower overlap reduction, or higher maintenance costs can push the payback period much longer.
The mistake: Using standard 5-10m GPS accuracy for applications where centimeter precision prevents accidents or damage.
Illustrative scenario 1: Autonomous Vineyard Sprayer A vineyard purchased an autonomous sprayer with standard GPS (5-10m accuracy). Vine rows were spaced 2.5 meters apart.
Failure mode: GPS error could cause the sprayer to:
- Drive over 12 rows of grapevines (15% of vineyard section)
- Spray adjacent properties (chemical trespass lawsuit)
- Get stuck in irrigation ditch (not detected, 8m off course)
Modeled impact: $85,000 in destroyed vines, $22,000 legal settlement, and $8,000 equipment recovery.
Root cause: Project manager didn’t understand GPS accuracy vs. row spacing. Assumed “GPS-guided” was sufficient without checking precision spec.
What RTK would have prevented: 2cm accuracy keeps sprayer centered in 2.5m rows with 1.23m clearance on each side. Even with 10cm error (degraded RTK), still 1.15m clearance—safe.
Illustrative scenario 2: Autonomous Delivery Robot A campus delivery robot used standard smartphone GPS (5-10m accuracy) to navigate sidewalks (1.5m wide).
Failure mode: GPS error could cause the robot to:
- Drive onto grass (got stuck 8 times in first week)
- Block wheelchair access ramps (ADA complaint)
- Enter vehicle traffic lane (3m GPS error made it think it was on parallel sidewalk)
Modeled impact: Program suspended after 2 weeks, with $120,000 in development effort stranded.
The fix: Sensor fusion
- Standard GPS for coarse positioning (which campus area?)
- Computer vision for sidewalk edge detection (stay within 1.5m path)
- Lidar for obstacle avoidance (don’t hit people)
- Lesson: Safety-critical navigation requires GPS PLUS other sensors, never GPS alone
Warning signs your project needs better GPS:
- Clearance < 10× GPS accuracy: Rows 50cm apart with 5m GPS = 10× too imprecise
- Damage risk: Off-course = crop destruction, property damage, safety hazard
- No backup sensors: GPS is only navigation input (no vision, lidar, lane detection)
Decision rule:
- Clearance ≥ 20× GPS error → Standard GPS OK (10m GPS for 200m-wide paths)
- Clearance 5-20× GPS error → Need DGPS or sensor fusion (5m GPS + vision)
- Clearance < 5× GPS error → Need RTK GPS (2cm GPS for 50cm rows = 25× margin)
GPS accuracy is not a single number — it varies from 3-5 meters (open sky, good geometry) to 50+ meters (urban canyons, poor GDOP). Enhancement technologies (DGPS, RTK, PPP) can improve accuracy to sub-centimeter levels but at significantly higher cost and complexity. The most important design decision is matching technology to requirements: over-engineering wastes money (RTK for fleet tracking), while under-engineering causes failures (smartphone GPS for autonomous farming). Always calculate the UERE error budget for your specific deployment environment.
GPS accuracy is like giving directions — sometimes “near the school” is good enough, but sometimes you need “third desk from the left!”
66.8.1 Accuracy Challenge
The Sensor Squad entered the “Location Olympics” with three different challenges, each needing a different level of accuracy!
Challenge 1: Find the Pizza Delivery Truck “Which street is the truck on?” asked Temperature Terry. Regular GPS said “Main Street, near the park.” That was close enough! The pizza was only a few minutes away. “5-10 meter accuracy is PLENTY for tracking delivery trucks,” Sammy announced. GOLD MEDAL!
Challenge 2: Guide the Robot Farmer “The robot tractor needs to plant seeds in rows that are only 50 centimeters apart!” said the microcontroller. Regular GPS said “somewhere around here” (5-10 meters off). The tractor ran over three rows of plants! DISASTER!
“We need RTK GPS!” said the battery. They set up a special base station that sent correction signals. Now the tractor knew its position within 2 CENTIMETERS. The rows were perfectly straight! GOLD MEDAL!
Challenge 3: Find the Lost Cat Indoors “Where is Whiskers in the house?” asked the LED. GPS said… nothing. Zero signal indoors!
“GPS doesn’t work inside buildings!” explained Sammy. “The satellite signals are too weak to go through walls. We need BLUETOOTH BEACONS instead!” They put tiny beacons in each room, and now they could tell Whiskers was in the kitchen (probably near the food bowl). GOLD MEDAL!
“Different accuracy for different jobs!” said the Sensor Squad. “Don’t use a ruler when you need a microscope, and don’t use a microscope when you just need a ruler!”
66.8.2 Key Words for Kids
| Word | What It Means |
|---|---|
| Error Budget | All the little mistakes that add up to make GPS less accurate (like adding up all the small delays) |
| DGPS | A helper station on the ground that fixes some GPS errors to make it more accurate |
| RTK GPS | Super-precise GPS that knows your position within 2 centimeters (width of a thumb!) |
| GDOP | How well-spread the satellites are in the sky — spread out is good, bunched together is bad |
How this chapter connects to other IoT concepts:
Builds on: GPS and Outdoor Positioning provides the fundamental GPS architecture and error sources that enhancement technologies correct. Applied in: Precision Agriculture autonomous tractors use RTK GPS for centimeter-level row guidance. Contrasts with: Indoor Positioning Technologies which use entirely different technologies (BLE, UWB) where GPS fails. Infrastructure: Fading and RF Interference shows how propagation effects and multipath degrade radio measurements, including GNSS signals. Cost trade-off: GPS accuracy selection balancing accuracy requirements against RTK infrastructure investment.
Related topics for deeper exploration:
Advanced GPS Techniques: Carrier phase, assisted-GPS, and multi-frequency techniques behind enhanced fixes. Differential Correction Services: Commercial NTRIP services (like Trimble RTX) that eliminate base station deployment. Agricultural Robotics: How autonomous tractors integrate RTK GPS with computer vision for obstacle avoidance. Precision Requirements: How centimeter-level accuracy requirements drive RTK selection. Kalman Filtering: Mathematical framework for fusing GPS with IMU to maintain accuracy during signal loss.
Hands-on exercises to understand GPS accuracy and enhancement:
Before collecting phone measurements, work through a signed-noise thought experiment that separates satellite geometry, receiver clock bias, and per-satellite measurement error. Compare the residuals you would expect with GDOP/PDOP: a rank-four solution can exist while poor geometry still amplifies small pseudorange errors.
Treat those expected residuals as simulated evidence, not as a claim about a particular receiver. The field exercise should still record the actual sky view, fix status, reported accuracy, and repeat scatter.
66.8.3 Calculate GPS Error Budget
Use a smartphone GPS app with satellite details:
- Record visible satellite count, HDOP, and reported accuracy for 10 different locations
- Note environment: open sky, urban canyon, indoors, under trees
- Calculate predicted error: UERE × GDOP (assume UERE = 6m from chapter)
- Compare predicted vs. reported accuracy
What to observe: Open sky GDOP ≈ 1.5-2.0 (good), urban canyon GDOP ≈ 4-8 (poor). Notice how accuracy degrades with fewer satellites and higher GDOP. Indoors, you may get NO fix or GDOP > 10.
66.8.4 Exercise 2: Simulate DGPS Correction
Visit GPS Visualizer or similar tool:
Record GPS coordinates at one fixed location for 30 minutes (log every 10 seconds). Plot the scatter of positions—you’ll see 5-15 meter spread. Calculate the centroid (average lat/lon)—this simulates a base station’s known position. Apply differential correction: offset = measured - centroid, then correct each point. Plot corrected positions.
What to observe: Uncorrected scatter is 5-15m diameter. After differential correction, scatter reduces to 1-3m. This demonstrates how DGPS removes common-mode errors.
66.8.5 RTK Precision Requirements
Design a scenario requiring RTK:
Crop rows are 50cm apart (standard corn/soybean). Tractor width is 3 meters, so it must stay within 25cm of row centerline. Standard GPS accuracy is 5-10 meters (100× too large!). Calculate: What accuracy margin do you need? Row width: 50cm. Tractor must not exceed 25cm deviation. Safety margin: ±5cm. Required accuracy: ±20cm worst case.
What to observe: RTK GPS delivers 1-2cm accuracy, providing 10× safety margin. DGPS (0.5-1m) would still damage crops. Only RTK meets agricultural precision needs.
66.9 What’s Next
Continue to Indoor Positioning Technologies to learn about Wi-Fi fingerprinting, BLE beacons, UWB, and sensor fusion techniques for positioning where GPS signals cannot reach.
| Previous | Up | Next |
|---|---|---|
| GPS and Outdoor Positioning | UX Design | Indoor Positioning Technologies |
