A final-year engineering project at Stellenbosch University examined how a home energy system can combine live inverter data, weather forecasts, and load-shedding schedules on one local dashboard. The result was a Raspberry Pi-based Home Energy Management System that monitors household energy flow, predicts demand, and adjusts battery charging and selected loads ahead of outages.

A final-year engineering project at Stellenbosch University explored how a smartphone can continue to function for basic communication when cellular or Wi-Fi networks are unavailable. The result was a two-part off-grid system that combines a mobile app with a compact LoRa device for messaging, location sharing, friend-finding, and low-resolution image transfer.

Strong turnout and engagement despite rainy weather Improved flow and experience with upgraded facilities

This research explores how model-free reinforcement learning can be used to improve customer engagement in e-commerce by optimising the timing of e-marketing communications. Focusing on open rates as a key performance metric, the study highlights how personalised delivery timing—based on individual customer behaviour—can significantly increase the likelihood of interaction. By developing and testing advanced reinforcement learning frameworks, including innovations like Targeted Adaptive Exploration, the study demonstrates how businesses can move beyond traditional marketing strategies toward more adaptive, data-driven approaches. The findings show strong potential for improving open rates, enhancing customer experience, and driving better overall marketing performance.

This research presents the development of a smart monitoring system for solar PV plants that improves efficiency, reliability, and maintenance through secure, real-time data insights. By combining AES-encrypted LoRa wireless sensor networks with smart field nodes, a scalable gateway, and an intuitive web interface, the system enables accurate performance tracking and early fault detection at panel level. The solution addresses key operational challenges by reducing maintenance costs, improving fault visibility, and supporting scalable deployment across large installations. With strong real-world performance and a modular design approach, the project lays a solid foundation for more efficient, secure, and future-ready solar energy management systems.

This research introduces Open-LACU, a novel machine learning framework designed to improve classification in real-world environments where data is often unlabelled and new categories frequently emerge. By combining semi-supervised learning with novelty detection, the approach enables models to classify known categories, identify previously unseen patterns within training data, and detect entirely new categories during deployment. Through experimental validation, the study demonstrates how Open-LACU can reduce reliance on costly labelled data while improving the safety and adaptability of classification systems. The framework represents a significant step toward more flexible, real-world-ready AI systems capable of handling uncertainty and continuous data evolution.

This research presents the development and validation of an electro-mechanical docking mechanism designed for CubeSats, addressing the unique challenges of small satellite coordination in space. By combining electromagnetic control, precise pose estimation, and a redesigned docking adapter, the system enables reliable, fuel-free docking between nanosatellites. Extensive simulation and experimental testing confirm the system’s accuracy, robustness, and readiness for real-world application in the DockSat mission. The project demonstrates a scalable solution that could unlock new capabilities in satellite servicing, in-space assembly, and collaborative CubeSat missions, advancing the future of space exploration.

This research explores how existing renewable energy inverters can be repurposed to improve voltage stability and reduce phase imbalance in rural South African power networks. Focusing on single-wire earth-return systems, the study highlights how uneven load distribution often leads to voltage unbalance beyond acceptable standards. By developing a dual-loop control system using MATLAB Simulink, the research demonstrates how inverters can actively regulate voltage through reactive and active power adjustments. Both simulation and real-world testing show that the approach effectively reduces voltage unbalance while maintaining stable grid performance. The findings present a practical, cost-effective solution to enhance electricity quality in rural areas by transforming passive renewable infrastructure into active grid-support tools.

This research explores how real-world objects can be mapped into virtual reality using 6DoF pose estimation and deep learning to create a more immersive, tactile VR experience. By combining convolutional neural networks with low-cost cameras and real-time tracking, the system allows users to see and physically interact with the actual objects they hold inside a virtual environment—bridging the gap between visual and physical perception. Through testing on various objects and comparing different tracking methods, the study demonstrates strong accuracy and practical feasibility, while also identifying key challenges such as object symmetry, occlusion, and data limitations. The findings highlight a scalable, cost-effective approach to enhancing VR immersion, with applications in education, training, and simulation.

Seven postgraduate engineering students will represent South Africa at the global Mission Idea Contest in Tokyo with their innovative SLINQI satellite concept. Designed as a practical, real-world solution, SLINQI proposes a pair of small satellites positioned on the far side of the moon to capture radio-based images of space, enabling the detection of cosmic signals that cannot be observed from Earth. The competition offers the team an opportunity to showcase South African engineering talent on an international stage while benchmarking their skills against global peers. Beyond the event, the team aims to bring back new knowledge and inspiration, contributing to the growth of local expertise in space science and engineering.