Robotics

A typical week in Year 2 might look like this: Monday starts with a robot kinematics and dynamics lecture—you're deriving the forward kinematics of a 6-DOF industrial robot arm using Denavit-Hartenberg parameters, building transformation matrices that map joint angles to end-effector position and orientation in 3D space. The maths is heavy with matrix algebra and trigonometry, but when you run your equations in MATLAB and see a simulated robot arm move to the exact position you calculated, the abstraction becomes satisfying. After lunch, you have a mechatronics lab where you're programming an Arduino-based mobile robot to navigate a maze using ultrasonic sensors and a PID control algorithm—tuning the proportional, integral, and derivative gains until the robot follows the wall smoothly without oscillating or crashing.

Tuesday brings a machine learning lecture on convolutional neural networks—how convolution layers extract features from images, pooling reduces dimensionality, and why dropout prevents overfitting. Your practical session has you training a CNN in PyTorch to classify objects captured by a robot-mounted camera, achieving 94% accuracy on a custom dataset you collected by driving the robot around the lab and labelling images. Wednesday is your heaviest day: a control systems lecture on state-space methods—modelling a quadrotor drone as a multi-input multi-output (MIMO) system, designing an LQR controller, and simulating the response in Simulink—followed by your group design project. Your team of four is building an autonomous mobile robot that must navigate an unknown environment, identify coloured targets, and manipulate objects. Today you're integrating the SLAM (Simultaneous Localisation and Mapping) algorithm with the path planner, and discovering that the LiDAR-based map drifts over time unless you add loop closure detection.

Thursday opens with a computer vision lecture on stereo vision and depth estimation—epipolar geometry, disparity maps, and how two cameras can reconstruct 3D structure from 2D images. The afternoon is an actuator and sensor systems lab where you characterise a brushless DC motor's torque-speed curve, interface an IMU (inertial measurement unit) over SPI, and fuse accelerometer and gyroscope data using a complementary filter to estimate orientation. Friday is lighter: a seminar on ethical and societal implications of autonomous systems—liability in autonomous vehicle accidents, algorithmic bias in AI decision-making, and the impact of automation on employment—followed by free time most students use for ROS programming, training neural networks on GPU clusters, or assembling their robot for the upcoming design competition. Weekends can be consumed by project work, especially as the competition deadline approaches, but building a machine that perceives, thinks, and acts in the physical world is one of the most rewarding challenges in all of engineering.