Partnerships

Familiar ground: Ageing gracefully in places we know

May 2026 | CDE News Research at CDE

Follow CDE

PDF Download

From drone swarms that track evasive targets to ultralight lattices that gobble up electromagnetic noise, collaborative research between CDE and Temasek Laboratories is producing technologies that translate ingenuity from the lab to the field.

Researchers at the College of Design and Engineering (CDE), National University of Singapore (NUS), are pushing the boundaries of autonomous robotics, aerospace guidance and advanced materials — fields in which progress often hinges on access to specialised infrastructure and domain expertise.

Working alongside Temasek Laboratories at NUS, a joint venture between the Ministry of Defence and the university with deep capabilities in aeronautical sciences, electromagnetics and information security, they are turning ambitious ideas into demonstrable systems. Three recent collaborations show what happens when engineering meets opportunity.

Eyes in the sky

Drone swarms — groups of autonomous drones that collaborate towards a shared goal — have a growing list of civilian uses, from environmental monitoring to search-and-rescue. They also play a vital role in military missions. But coordination is fiendishly hard. As the number of drones in a system rises, the volume of possible states and actions grows exponentially — a problem roboticists call the “curse of dimensionality.”

Assistant Professor Guillaume Sartoretti, from the Department of Mechanical Engineering at CDE has spent years cracking this curse. His Multi-Agent Robotic Motion (MARMot) Lab uses deep reinforcement learning to train individual agents to make local decisions that collectively yield large-scale coordination. Through a partnership with Temasek Laboratories, Asst Prof Sartoretti’s team has turned these principles into working drone systems.

Their trajectory began with autonomous exploration of unknown areas, first with a single robot and then with multiple, before advancing to adversarial search: scenarios where the targets actively evade detection. The result was ViPER (visibility-based pursuit evasion via reinforcement learning), which tackles the famous “art gallery problem” — the question of how many guards are needed to surveil an irregularly shaped space — by deploying drone swarms that must clear an entire environment while preventing targets from slipping back into previously searched zones. Using distributed deep reinforcement learning, the drones cooperate in a fully decentralised manner, each acting on local information alone. The system has been validated in both simulations and physical robot demonstrations.

Assistant Professor Guillaume Sartoretti’s team uses deep reinforcement learning to train individual agents to make local decisions that collectively yield large-scale coordination.

A parallel challenge is persistent surveillance: maintaining continuous awareness of targets that keep moving. COMPASS, the team’s multi-agent reinforcement learning framework presented at the IEEE International Symposium on Multi-Robot and Multi-Agent Systems (MRS 2025), addresses this challenge. Where ViPER guarantees detection of evasive targets, COMPASS keeps tabs on dynamic ones over extended periods, pairing Gaussian Processes for uncertainty estimation with a spatio-temporal attention network that enables each drone to reason about where information will be most valuable next. The framework was a finalist for the Best Paper Award at the symposium, held in December 2025 in Singapore.

“The fundamental insight is the same across both systems,” adds Asst Prof Sartoretti. “If each agent is equipped with the right learning architecture, coordination emerges naturally, and it can scale, without a central planner directing every single drone.”

“If each agent is equipped with the right learning architecture, coordination emerges naturally, and it can scale, without a central planner directing every single drone.”

Attitude adjustment: sticking the landing

In 2015, SpaceX turned science fiction into science reality by autonomously landing the booster of its Falcon 9 rocket — the vehicle’s most expensive component, which accounts for more than half of its total manufacturing cost. The feat proved that reusable rockets could dramatically cut the cost of reaching orbit. Companies including Blue Origin have since followed suit, but the underlying engineering challenge remains: guiding a rocket to a precise touchdown without human intervention, a puzzle known as the powered descent guidance (PDG) problem.

Early solutions modelled the rocket in just three degrees of freedom (DOF), accounting for translational movement along the x-, y- and z-axes. While flight tests validated the approach, realistic rocket behaviour demands six degrees of freedom, adding the rotational dynamics of pitch, yaw, and roll — what engineers call the vehicle’s “attitude.” Solving the PDG problem in this fuller model and under various realistic operational physical limits has proven far harder, and most methods remain confined to simulations.

Assistant Professor Zhao Lin, from the Department of Electrical and Computer Engineering at CDE, is working to close that gap. In collaboration with Senior Research Scientist Dr Rodney Teo, who heads the Swarm / Autonomy Group at Temasek Laboratories, Asst Prof Lin is developing a complete 6-DOF rocket model equipped with a thrust vector control (TVC) system — the mechanism by which a rocket adjusts the direction of its engine thrust to govern orientation and angular velocity.

The research spans trajectory planning algorithms, which chart the rocket’s optimal descent path by converting a complex non convex optimisation problem into a sequence of tractable convex sub-problems, and tracking control algorithms that keep the vehicle locked onto that path in real time. The project will progress from software simulations to hardware tests — first using a proxy electrical thruster model, then a hybrid propellant system, with Temasek Laboratories’ propulsion team providing critical support at each stage.

The culmination will be a tethered grasshopper manoeuvre: vertical takeoff, horizontal traversal, powered descent and vertical landing — a controlled proving ground that bridges simulation and free flight. If the thruster performs the sequence autonomously, safely and precisely, it will be a giant leap for operationally reusable rocket vehicles.

“The constrained 6-DOF planning and control problem is where theory meets the real physics of rocket flight,” adds Asst Prof Lin. “Our goal is not just to solve it in simulation, but to demonstrate it on hardware — that’s where the true test lies.”

Assistant Professor Zhao Lin is developing a complete 6-DOF rocket model equipped with a thrust vector control system, which enables the rocket to adjust the direction of its engine thrust to govern orientation and angular velocity.

“Our goal is not just to solve it in simulation, but to demonstrate it on hardware — that’s where the true test lies.”

A lattice shield for improved connectivity

Humming and buzzing with connectivity, Singapore ranks among the world’s smartest cities. And with that, the number of Internet of Things (IoT) connections in the country has skyrocketed in recent times, and is projected to grow substantially in years to come. Every one of those connections relies on electromagnetic signals — and as the wireless environment grows denser, so does the risk of electromagnetic interference (EMI), a form of pollution that can disrupt medical devices, military infrastructure and potentially human health.

Modern electronics incorporate shielding to limit interference and maintain electromagnetic compatibility — a device’s ability to function without causing or suffering disruption. But as IoT proliferates, demand for lighter, more customisable shielding is outpacing what conventional materials can offer.

Associate Professor Yan Wentao and his team at the Department of Mechanical Engineering, CDE, took on this challenge in collaboration with Senior Research Scientist Dr Yang Yong’s Electromagnetic Materials Group at Temasek Laboratories. Funded through a Temasek Laboratories seed project, the researchers turned to fused deposition modelling — a 3D-printing technique in which melted plastic filament is extruded layer by layer to form complex structures, enabling rapid scalability and customisation.

Associate Professor Yan Wentao and his team developed ultralight 3D-printed lattices capable of absorbing broadband microwaves across a wide frequency range.

Using conductive polylactic acid rather than metal, the team created lattice structures capable of absorbing broadband microwaves across a wide frequency range. The lattice architecture achieves sufficient electrical conductivity for effective EMI absorption while maintaining a density as low as 0.1g cm-³. That ultralight profile means the structures add virtually nothing to
the weight of the systems they protect — a critical advantage for aerospace applications.

Commercially available alternatives, such as microwave-absorbing foam, lack the same degree of design flexibility. The 3D-printed lattices can be tailored to exceed 90% absorption efficiency for broadband microwaves while retaining mechanical integrity, making them suitable for weight-sensitive platforms where every gram counts.

“The beauty of this approach is its adaptability,” says Assoc Prof Yan. “We can tune the lattice geometry to target specific frequency ranges, and the manufacturing process scales easily, from a laboratory prototype to a production run.”

Common ground

These projects are a snapshot of the collaborative research carried out by Temasek Laboratories and CDE, who continue to seed new programmes across autonomy, aerospace and materials science, with many more in the pipeline.

“Our mission is to cultivate strong, multidisciplinary research teams to bring together theoretical possibilities, practical engineering insight, and a spirit of innovation,” says Associate Professor Vincent Tan, Director of Temasek Laboratories at NUS and a Dean’s Chair Professor at the Department of Mechanical Engineering, CDE.

“Our mission is to cultivate strong, multidisciplinary research teams to bring together theoretical possibilities, practical engineering insight, and a spirit of innovation.”

“Collaboration is central to our strategy. We partner closely within NUS, as well as with local and global institutions. Working in close tandem with the defence community, we bridge gaps between foundational research and real-world defence applications. This synergy helps align our research with national technological needs and enhances application impact.”