Neal Batchelor's
Portfolio

The Market Challenge

During the COVID-19 pandemic, UK pet adoptions surged. As owners returned to the office, dogs were left isolated for longer periods, leading to disruptive behaviours. Existing pulling toys on the market all require a human to be on the other end, providing the resistance for the dog to counter. There is no available self-automated tug-of-war system on the market. The aim was to design a mechanically adjustable device capable of independent, prolonged engagement while safely withstanding the dynamic loads of a lunging dog.

Mechanical Solution

The 'Tug of Paw' is a wall-mounted mechanical system driven by a 12mm En8 steel shaft. To allow for independent play, the system utilises a torsional clock spring that stores potential energy when the dog pulls the dynamic climbing rope, automatically reeling the rope back in upon release.

To accommodate dogs of varying strengths, pulling resistance is generated by a mechanical friction brake clamped over a custom steel disc. Crucially, a CSK12PP Sprag Clutch (one-way bearing) is integrated into the shaft. This ensures the friction brake only applies resistance when the dog pulls, allowing the spring to retract the rope freely without fighting the brake's friction.

FEA & Iterative Optimisation

Following the Pahl and Beitz design methodology, rigorous FEA was conducted based on a 2000N dynamic load (simulating a 40kg dog lunging at 5m/s).

Brake Disc: To ensure the spring could successfully retract the rope, the rotational inertia of the shaft components had to be minimised. Through topological optimisation, the brake disc's mass was reduced by 40% (550g to 332g), decreasing its moment of inertia by 37% while maintaining a minimum safety factor of 4.6.

Modular Hook: A custom carabiner-style hook was developed to allow owners to swap out toys. Initial FEA identified peak stresses of 26 GPa at the mounting hole. Through shape optimisation, material was removed from low-stress regions and load distribution was improved, reducing base stress by 87%.

The Engineering Challenge

A two-wheeled self-balancing robot is an inherently unstable system — without continuous active correction it will topple immediately. The challenge was to program a fixed-hardware platform to sustain autonomous balance on carpet using MATLAB R2024a and Simulink, against a defined set of customer requirements including a 15-second minimum balance duration and a child-appropriate personality for University open days.

Control System Solution

A closed-loop control system was built in Simulink using the robot's onboard IMU. Raw pitch data was converted to degrees via a calibrated gain block and compared against a desired setpoint adjustable through an onboard potentiometer. A PID controller block computed corrective PWM motor commands from the resulting pitch error signal, with a saturation block clamping output to the range the motor driver accepts.

Gains were tuned using a Ziegler–Nichols-informed iterative method — proportional gain was raised until sustained oscillation appeared, derivative action was then increased to damp it, and integral action was finally introduced to eliminate steady-state error. This process was repeated across multiple test runs on both hard floor and carpet until stable, long-duration autonomous balance was achieved on all tested surfaces.

The Engineering Challenge

Fixed photovoltaic panels only reach peak efficiency when the sun is perpendicular to their surface — a condition that occurs briefly each day. The aim of this project was to design, build, and test a fully autonomous two-degree-of-freedom solar tracker that could outperform a static panel. It had to demonstrate dual-axis rotation, closed-loop feedback control and safe power electronics.

Electro-Mechanical Solution

The tracker uses two servomotors driven through a 1:1 spur-gear transmission to achieve independent yaw (azimuth) and pitch (elevation) control. A needle-roller bearing carries the rotating platform's structural weight, ensuring the azimuth servo only handles rotational actuation rather than structural loads — critical given its 27 Ncm torque limit.

Four LDRs mounted in an X-formation with physical shading walls feed a closed-loop PD controller running on an Arduino R4 WiFi. The controller compares opposing LDR voltage pairs to determine light imbalance and commands servo corrections.

Because servos are limited to 180°, the tracker was programmed to flip at either boundary to continue following the light — transposing the LDRs, inverting the error signal, and driving the panel away from the source. A custom inversion algorithm detects this, negates the steering logic for both axes, and restores correct closed-loop behaviour.

Testing & Results

The system was validated under a simulated sun-path test, with a flashlight swept in an arc to replicate solar movement. PV voltage from both the tracking and a static reference panel were logged to an SD card.