Hi, I'm Mark

Robotics perception engineer — SLAM, 3D mapping, and sensor fusion

I've spent the last decade or so building autonomous robots — from research platforms to systems now deployed commercially. My work runs from perception software through to mechanical design and embedded hardware.

Current Work

Robotics Software Engineer

At SEQSENSE, I develop perception and mapping software for autonomous robots: SQ-2, a security robot deployed as a fleet across airports, offices, and public buildings in Japan, and FORRO, a delivery robot co-developed with Kawasaki Heavy Industries.

SQ-2 at Narita Airport — SQ-2 performing a patrol at Narita Airport (Tokyo)

Both robots use a rotating turntable with three 2D LiDARs for navigation and wide-angle cameras for 360-degree vision. They transmit real-time data to a web platform for monitoring and can autonomously dock for charging and operate elevators for multi-floor operations.

My work spans online lidar-based 3D mapping that runs in real time on desktop, in the cloud, and on the robot itself; 3D object detection and tracking from lidar and camera, with detection models for people and other robots trained in PyTorch and OpenVINO; and Gazebo simulation that mirrors robot behaviour, with robot models built from CAD. The video shows a prototype interface I developed for real-time SLAM mapping using ROS and C++. Users could view maps as they were built, with scan quality indicated in green. The interface was designed for tablet/smartphone display mounted to a PlayStation controller.

Web-based interface for SLAM mapping

Agricultural Robotics R&D

Kiwifruit Automation Project

As a post-doctoral researcher with the University of Waikato and University of Auckland, I worked on robotic systems for kiwifruit operations. The project involved collaboration between universities, Plant and Food Research, and Robotics Plus Ltd, funded by the Ministry of Business, Innovation and Employment. The research team developed autonomous systems for pollination, harvesting, and platform mobility, though the technology ultimately proved economically challenging for commercial deployment.

Multi-purpose platform in kiwifruit orchard — The multi-purpose platform in a kiwifruit orchard

Autonomous Platform Development

The multi-purpose platform served as a mobile base for harvesting and pollinating systems. This hybrid petrol-electric vehicle achieved autonomous row-following using LiDAR-based orchard structure sensing. Six electric motor/gearbox units provided full electric drive capability while supplying AC and DC power to mounted systems. The platform supported 1 tonne payload capacity with additional fruit storage between rear wheels.

Robotics platform being driven around a kiwifruit orchard

Harvesting System

The harvesting system identified kiwifruit in 3D space using stereo cameras and neural networks. It determined optimal picking sequences to minimize damage to neighboring fruit before directing robotic arms to collect fruit. The integrated approach combined computer vision, machine learning, and precision robotics for orchard operations.

Kiwifruit harvesting system in operation

Pollination System

The pollination system used the same camera and neural network infrastructure to identify flowers in 3D space. It calculated pollen solution trajectory timing to account for vehicle motion, ensuring accurate flower targeting while driving. The system enabled autonomous pollination operations across orchard rows.

Selected Publications

See all publications ▸

Open Source & Tools

Container-Magic — reproducible containerised development environments (my active flagship tool).
dmenu-extended — an extensible launcher menu for Linux.
Blender PCD I/O — point-cloud import/export for Blender.
Calibration checkerboards — print-ready targets, widely used in robotics and computer vision.