Competition Guide

Day 1 — Testing Day (Thu 30 July)

Event Description
Technical inspection Dimensions, weight, mandated sensors, battery safety, one live reading. Pass/fail — 30 min to rectify a failure
Measurement run 8 minutes; operator stays in the designated zone; stuck robot is placed at arena centre by officials
Data review Judges compare readings against reference values immediately after the run
Robot assessment Cats 3/4/5 judged + 3-minute innovation pitch + questions

Day 1 Morning Preparation

  1. Full LiPo charge (spare charged too).
  2. Flash latest firmware; verify boot into RC mode (BOOT held → LED on).
  3. RobotAP visible; 192.168.4.1 loads on phone.
  4. Serial monitor on laptop — confirm sample lines print with units.
  5. Test both sensors with known references; re-check calibration constants.
  6. Drive test on a rough surface — confirm torque is adequate.
  7. Confirm GET /data returns valid JSON and the visualiser accepts it.
  8. Walk the Inspection Checklist.

Day 1 — 3-Minute Innovation Pitch

Segment Duration Key message
Seesaw arm 35s "One servo deploys two sensors — simpler, lighter, fewer failure modes than two actuators."
Pond safety 35s "The colour sensor looks ahead of the wheels; the robot samples, then reverses away — wheels never touch the recessed water slots."
Dual-mode firmware 35s "The robot is its own infrastructure: WiFi AP, RC page served from flash, autonomous wander as a stretch mode."
Engineering depth 35s "Two I2C buses resolve a sensor address clash without a multiplexer; the turbidity sensor runs through a calibrated voltage divider."
Data pipeline 40s "Every reading streams to serial and JSON; the file drag-drops into our offline visualiser with arena replay."

Only claim what is on the robot. Judges ask follow-ups — the rulebook scores teams down for components they cannot explain.


Day 2 — Closing Night Case Competition (40 pts)

Phase Details
Problem reveal Scenario given at the start of the night — unknown in advance, no single correct answer
Preparation 1 hour to build a response from Day 1 data
Presentation Up to 3 minutes + Q&A; multiple members must contribute (teamwork is scored)

1-Hour Preparation Plan

Time Task
0–5 min Open the saved results.json in the visualiser; verify all samples render
5–15 min Read the scenario; pick which readings are most relevant
15–30 min Build the argument: data → environmental conclusion → recommendation
30–45 min Draft the 3-min script; choose visualiser views for the projector
45–60 min Practice run + Q&A rehearsal; assign speaking segments

Argument Template

Opening (30s):
"Our robot collected [N] readings across four terrain types.
The data shows [key finding]."

Data per terrain (30s × 4):
"In [terrain], turbidity was [X] NTU — [clean / moderate / turbid].
Soil moisture was [Y]% — [dry / healthy / waterlogged]."

Recommendation (30s):
"We recommend [specific action] because [data-backed reason].
The greatest concern is [terrain] due to [specific reading]."

Limitations (20s):
"Point-in-time snapshot; single run; ±10% sensor band;
sector labels are operator-tagged; path is an estimate."

Closing (10s):
"Both sensors were calibrated against reference values before the run."

Acknowledging limitations is scored positively — tradeoff handling is an explicit Cat 6 criterion.


Q&A Preparation

Question Response approach
"How accurate are your sensors?" "Calibrated against reference samples before the run; the turbidity curve was refitted after adding the voltage divider. Target ±10% per Cat 1."
"Why remote control instead of autonomous?" "The rulebook permits RC. Our robot has an autonomous wander mode, but operator-driven sampling guarantees correct sector labels and eliminates navigation failure modes — accuracy is worth more points than autonomy."
"How do you avoid the water hazards?" "The colour sensor is mounted ahead of the front wheels. On detection the robot stops, samples with the arm, then reverses away — wheels never cross a slot."
"How do you know you didn't sample the same spot twice?" "In RC mode the operator tags each sample by sector against a 12-sample checklist. In autonomous mode a driving-time lockout plus a heading change after every sample prevents immediate re-detection."
"What are your data's limitations?" Name them proactively: single run, point-in-time, path is a speed-model estimate, ±10% band.