← Back to Projects
Autonomous Systems Hardware Prototyping Arduino FEEG2001

Autonomous
Specimen Rover

A fully autonomous ground rover capable of arena-wide self-localisation, squash ball collection, and sequential deposition — built within a £50 materials budget.

Year2025
ModuleFEEG2001
PlatformArduino UNO + MD25
StructureLaser-cut plywood
RoleTeam Lead
IR-9 autonomous rover — side view
±1 cm
Straight-line accuracy
post-PID (30–60 cm)
±0.8°
Rotational accuracy
at speed 5 (post-fix)
HC-SR04 ultrasonic
sensors
£31.60
Build cost
(within £50 budget)
Task A
Completed successfully
in demonstration

Self-navigating rover for arena-based sample collection

The IR-9 rover was designed and built for the FEEG2001 Systems Design and Computing module. Inspired by NASA's Perseverance and Spirit rovers, the brief required a fully autonomous vehicle capable of two sequential tasks inside a walled arena — with no GPS or external localisation. The rover needed to navigate, collect simulated rock samples (squash balls), and deposit them into two separate collection cups.

Led by Daniel Tombleson, the team of five took the project from concept sketches and cardboard mockups through to a laser-cut plywood prototype with full embedded software — delivered within a £50 materials budget.

Task A

Location & Navigation

Starting from an undisclosed position, the rover spins and compares paired rear ultrasonic readings to detect wall parallelism, then navigates autonomously to the target zone with a buzzer and LED confirmation on arrival.

45 sec time limit
Task B

Collection & Deposition

A servo-actuated retractable arm collects four squash balls in one sweep. A Ferris-wheel mechanism then dispenses two balls into each of two cups, with the rover navigating between them using encoder odometry.

100 sec time limit

System components & design decisions

The rover integrates encoder-based PID motion control, four-channel ultrasonic localisation, a servo-actuated collection arm, and a Ferris-wheel deposition mechanism — all driven by a single Arduino UNO communicating with an MD25 dual motor driver over I²C.

Microcontroller
Arduino UNO R3
Motor Driver
MD25 (I²C, dual-channel)
Drive Motors
Brushed DC with encoders
Sensors
4× HC-SR04 ultrasonic
Actuation
2× servo (scoop + wheel)
Power
12 V lead battery
Structure
6 mm laser-cut plywood
Moving parts
3D-printed PLA
Control law
PID deceleration
Comms protocol
I²C (Wire library)
Sensor library
NewPing
Max dimensions
350 × 400 × 400 mm
Sensor relocation: Rear-facing sensors were repositioned to exploit the HC-SR04's accurate <100 cm range for wall-parallelism detection. Sensors at longer ranges (150–200 cm) exhibited up to 32 cm error — constraining detection to the short-range band was the decisive reliability improvement.

Systematic bench testing & calibration

Bench testing covered straight-line travel, rotational accuracy, and sensor range across multiple speeds and distances. A float/integer casting bug in the wheel-diameter-to-turn-angle ratio was identified and corrected, reducing rotational drift to zero at operating speed.

Test conditionTargetMeasuredErrorResult
Straight-line — 30 cm, speed 5 (pre-PID)30.0 cm33.2 cm+3.2 cmOvershoot
Straight-line — 30 cm, speed 5 (post-PID)30.0 cm31.0 cm+1.0 cmPass
Straight-line — 60 cm, speed 5 (post-PID)60.0 cm61.0 cm+1.0 cmPass
Rotation — 360°, speed 5 (pre bug-fix)360°362°+2°Drift
Rotation — 90°, speed 5 (post bug-fix)90°90.8°+0.8°Pass
Sensor range ≤ 100 cmExact±0 cmReliable
Sensor range ≥ 150 cm153–170 cm+3 to −32 cmDegraded

Electronics integration & prototype construction

The rover body was constructed from laser-cut 6 mm plywood — net-cut into interlocking sections for precise assembly. All wiring was secured to prevent tangling with moving parts, and a fuse was placed in series with the main power switch to protect the Arduino and MD25 from overcurrent.

Electronics layout during integration testing
01 — Electronics during integration: Arduino UNO, MD25, breadboards, HC-SR04 sensors
Internal electronics bay
02 — Internal bay: Arduino, MD25 driver, LED display, and wiring harness
Front panel with ultrasonic sensors
03 — Rear ultrasonic sensors: paired HC-SR04s used for wall-parallelism detection
Rover in arena
04 — Rover frame and casing positioned inside the target zone during Task A

Outcomes

Results & reflections

Task A performance
Demonstrated successfully — rover self-localised, navigated to the target zone, and triggered buzzer/LED confirmation within the 45-second limit
Task B outcome
Ball collection and cup navigation completed correctly; deposition was hindered by a Ferris-wheel misalignment — addressable with higher-tolerance 3D-printed components
PID implementation
Proportional deceleration eliminated momentum-driven overshoot, achieving ±0 cm and ±0° accuracy — critical for reliable autonomous navigation without external reference
Budget performance
£31.60 material cost against a £50 limit — achieved through selective use of 3 mm vs 6 mm plywood and minimising 3D print time
Key learning
Sensor placement relative to operating range has a disproportionate impact on reliability — constraining detection to the <100 cm accurate band was the decisive design improvement
Commercial pathway
FEEG2006 commercialisation report identified the strongest application as an autonomous tennis ball collector — priced at £799.99 against competitor Tennibot at ≈£3,000