SPECTRUM
The Challenge
Name:
Stereo event-camera on miniature drone
Domain:
Sensor integration, computer vision, design and development of miniaturized drones, advanced camera systems
Challenge proposer:
Objectives:
SPECTRUM addressed AUTOASSESS Challenge 2: proving that stereo event-based cameras can enable autonomous inspection in confined, GPS-denied maritime spaces such as cargo holds, ballast tanks and engine rooms —environments that are dark, low-texture and hazardous for human surveyors. Conventional frame-based drones struggle here, and high-end inspection platforms are costly to adopt. The core challenge was integrating event-camera perception, event-based SLAM, 3D mapping and autonomous path planning onto a compact, affordable airframe, and demonstrating the full stack flying end-to-end without GPS. A further challenge was doing this at low cost, to lower the barrier for shipyards and classification bodies to evaluate the technology.
The Solution
Name:
SPECTRUM – Stereo-perception enabled compact robot for unmanned monitoring
Description:
SPECTRUM developed a compact, low-cost drone for autonomous inspection of confined, GPS-denied spaces inside vessels — cargo holds, ballast tanks, void spaces and engine rooms — where human entry carries real safety risk. Built on a 6-inch airframe (around €720 in hardware, excluding cameras), it pairs stereo event-based cameras with a fully onboard autonomy stack: an NVIDIA Jetson runs event-based SLAM, 3D mapping and A* path planning, while a PX4 flight controller handles stabilisation and safety. During the AUTOASSESS Assess phase the team integrated all subsystems and demonstrated end-to-end autonomous indoor flight.
Key Results Delivered:
SPECTRUM delivered a working 6-inch-class inspection drone (≈€720 in hardware, excluding cameras) that combines stereo event-based cameras with a fully onboard autonomy stack. An NVIDIA Jet-son runs event-based SLAM, 3D occupancy mapping and A* path planning; a PX4 flight controller handles stabilisation, with a LiDAR safety layer and MAVLink offboard control linking the two. All seven project milestones were achieved, including a successful integrated autonomous flight in a GPS-denied indoor space, with live event streams, pose estimates and 3D maps produced onboard.
Unique Value and Impact:
SPECTRUM is one of relatively few publicly documented demonstrations of stereo event-based SLAM driving a full onboard autonomy stack on a sub-€1k miniature drone in a GPS-denied environment. Its unique value is accessibility: a low-cost reference platform that lets shipyards, classification societies and inspection providers evaluate event-camera inspection technology without committing to expensive commercial systems. Reusable engineering outputs — a custom event-camera calibration procedure, an ESVO2 configuration set, and a PX4-and-Jetson offboard control architecture — lower the integration cost for the wider community. The longer-term impact is safer, cheaper robotic inspection of confined, hazardous vessel spaces that today require risky human entry.
Solution in action:
Picture a surveyor preparing to inspect the ballast tanks and cargo holds of a bulk carrier in dry dock — spaces that are dark, cramped, GPS-denied and hazardous to enter. Instead of sending a person in on ropes, the operator launches SPECTRUM, a 6-inch drone that fits through access hatches. Using its stereo event cameras, the drone perceives the compartment even in low light and low-texture conditions, building a live 3D map onboard. Its Jetson companion computer runs event-based SLAM and A* path planning while the PX4 flight controller maintains stable flight, with a LiDAR safety layer preventing collisions. Guided from a single operator console, the drone flies autonomously through the space, recording event streams, pose estimates and a 3D point-cloud map for offline review — all without GPS or external infrastructure.
The essential functionalities are onboard perception, autonomous navigation and 3D mapping in confined, GPS-denied environments. Potential users include shipyards, classification societies, inspection service providers and vessel owners seeking safer, faster and lower-cost surveys. Beyond ships — cargo holds, void spaces, double-bottom structures and engine-room internals — the same platform extends naturally to adjacent confined-space domains such as mine shafts, tunnels, storage facilities and large industrial process equipment.
The Solution Provider
CogniSensus builds technologies that allow better sensing, data modelling, visualisation and/or actuation, supporting the operation, decision- making or training in complex environments.
Key Team Members:
The AUTOASSESS Experience
Main contribution of the AUTOASSESS programme:
The AUTOASSESS Open Call gave CogniSensus the funding, focus and framework to take stereo event-camera inspection from concept to a working, integrated prototype within a structured six-month programme. The staged milestones — planning, development, and assessment — imposed a disciplined path from paper design to a flying, autonomous platform. Equally valuable was access to the AUTOASSESS consortium: partners spanning classification societies, shipping operators, robotics developers and academic groups working on event-based vision, providing a directly relevant network for validation, feedback and future collaboration. The programme also anchored the work in a concrete industrial need — safer, lower-cost inspection of confined maritime spaces — sharpening both the technical goals and the commercial pathway.
Testimonial
“Taking part in AUTOASSESS gave us the chance to prove something we believed but hadn’t yet built: that stereo event-based vision and a full autonomy stack could run on a compact, affordable drone in the kind of dark, GPS-denied spaces where inspection is hardest. The programme’s structure pushed us from design to a flying prototype in six months, and being part of the consortium connected us to the shipping, classification and robotics partners who understand the real problem. We came away with a working platform, a clear path forward and a much sharper view of where this technology can go.”
Vasileios Balntas, Project Lead