Collaborative autonomy · At the edge

I build efficient spatial intelligence for robots to perceive, navigate, and act together, envisioning a future defined by collaborative autonomy.

Miyuru Thathsara in a candid portrait
PhD Researcher @ NTU, Singapore Miyuru Thathsara

The vision

My vision is not one all-powerful robot. It is a collective of small, efficient machines that share spatial understanding, coordinate without a central point of failure, and create practical value for people.

Why collaboration matters

Collaboration turns limited robots into a capable collective.

Applications

Coordinated drones can divide large fields, share crop and weed observations, and redirect effort toward emerging hotspots. The EU-funded SAGA project explored this approach for faster weed mapping and more targeted field treatment.

Collaborative precision agriculture mission Three drones coordinate their scan paths over crop rows and share a detected crop stress location. SHARED FIELD MAP CROP STRESS

A cooperative fleet can divide bridges and rail corridors into inspection zones, capture complementary viewpoints, and synchronize findings. The EU Drones4Safety project developed a collaborative swarm system and tested it in bridge and railway inspection cases.

Collaborative bridge inspection mission Three drones inspect a bridge from different positions while combining observations of a structural anomaly. MULTI-VIEW ALIGNMENT DEFECT CONFIRMED

The August 2026 Nepal–Tibet glacier collapse triggered a catastrophic flood with severe loss of life. At Blatten in 2025, surveillance and hazard assessment enabled residents to evacuate nine days before a rock–ice avalanche destroyed much of the Swiss village; no resident died.

Blatten was not monitored by cameras alone. On the evacuation day, ground-based interferometric radar was installed in the valley to track movement; a camera installed on the glacier’s eastern moraine the day before the collapse captured its final acceleration. Across the far larger, remote Himalaya, collaborative drones could complement satellites and selected ground sensors by repeatedly mapping difficult slopes and sharing changes for earlier risk assessment.

Collaborative landslide monitoring mission Three drones scan a steep slope, identify active ground movement, and relay a shared early warning. LANDSLIDE SCENARIO · LIVE TERRAIN CHANGE ACTIVE MOVEMENT

From swarm to system

Focusing on one at a time.

A capable collective begins with efficient intelligence on every robot.

An agent selected from a collaborative swarm Several drones coordinate over an environment while a highlighted agent reveals its onboard processing stages.

Onboard processing

Parallel stages

Collaborative computation

I bet on two things at once.

Point to or tap each bet to see why

Follow the research on Google Scholar

Bet 01 · Camera vision

Visual localization for collaborative agents.

Separate views become one shared understanding of motion and place.

Simplified from peer-reviewed 2024 IEEE research on collaborative drones

Point to or tap a stage

Runs onboard every agent · useful observations move across the team

Bet 02 · FPGA-based SoC computing

FPGA-based systems-on-chip for low-volume, high-value applications.

Point to or tap a component

FPGA-based system-on-chip

Connect

Let’s make this
happen.

Let’s connect around collaborative robotics, visual localization, or efficient onboard computing.