An autonomous electric mower that uses LiDAR, GPS and an IMU to navigate and cut grass.
An autonomous mower that handled unfamiliar outdoor ground under its own navigation.
Stage reached
Field tested
Period
Co-founded venture
Sector
Robotics and autonomous vehicles
Organisation
Elite Robotics
The autonomous mower during outdoor testing.
My contribution
As co-founder and embedded engineer I led the product and integrated its sensors, computers and vehicle controls.
Embedded architecture and C++ / ROS integration
LiDAR, GPS and IMU sensor integration
Vehicle control, outdoor testing and demonstrations
The result
An autonomous mower tested outdoors, with a reusable foundation for other service vehicles.
Engagement
Autonomy platform engineering and systems integration
The problem
Boundary wire and random bounce was never going to be the product. The mower had to map outdoor terrain it had not seen, hold a position estimate under tree cover, notice obstacles that move, and plan complete coverage, all on compute small enough and cheap enough to sit on a service vehicle.
How it was approached
Use one mower as the proving ground for the entire autonomy stack rather than simulating around the hard parts.
Keep sensing, localisation, planning and vehicle control separable, so any one of them can be tested or replaced.
Make coordinate frames and timing visible before sending a single motion command.
Move from bench integration to outdoor runs and public demonstrations as fast as the vehicle allowed.
How far up the stack the work reached
The vehicle platform and drivetrain already existed. Elite Robotics built the sensing, compute, navigation and control layers around them.
Built
Integrated
Already there
Machine
L1
Electric mower platform, drivetrain and cutting deck.
Sensing
L2
M8 LiDAR, GPS and IMU acquisition with aligned frames and timing.
Embedded
L3
Raspberry Pi and NVIDIA Jetson interfaces for power, sensing and motion.
Compute
L4
C++ ROS nodes, message contracts and launch configuration.
Navigation
L5
Localisation, multi-target obstacle tracking and coverage planning.
Control
L6
Motion commands driven into the RoboClaw vehicle controls.
Engineering load
Mechanical
Electrical
Embedded
Software
Operations
What was built
01
Onboard vehicle software
The compute, the sensors and the vehicle, speaking the same language.
Raspberry Pi and NVIDIA Jetson integration
ROS nodes, message contracts and launch configuration
Hardware interfaces for power, sensing and motion
Diagnostics and teleoperation paths for field debugging
02
Localisation and navigation
Knowing where the machine is before telling it where to go.
LiDAR, GPS and IMU acquisition with filtered fusion
Coordinate frame and timing alignment across sensors
Probabilistic sensor fusion and multi-target tracking
Coverage and obstacle-aware path planning
03
Outdoor field testing
Grass, slopes and weather, which is where bench assumptions die.
Estimated vehicle state wired through to movement commands
Repeated outdoor runs with tuning between each
Public demonstrations on a working golf course
Failures isolated to sensing, planning or control rather than guessed at
What changed
22 repositories
A real engineering estate
Navigation, guidance, target tracking, LiDAR, GPS, IMU filtering, motor control, diagnostics, teleoperation, embedded firmware and cloud services, each versioned separately.
Field tested
Autonomous on grass
The platform went from bench integration to a mower cutting grass autonomously under its own navigation stack.
NSW MVP grant
Government backing
Elite Robotics received a NSW Government Minimum Viable Product grant.
Slingshot
Accelerator and venture funding
The company completed the Slingshot Accelerator and received backing through its venture fund.
Technical stack
C++
ROS
NVIDIA Jetson
Raspberry Pi
LiDAR
GPS
IMU
MATLAB
Capabilities applied
Robotics software
Sensor fusion
Motion planning
Embedded compute
From the project
The vehicle platform beneath the bodywork.Field testing the integrated vehicle.
Watch it run
Navigation and sensor fusion
Range-sensor data is filtered into terrain and object clusters while the system estimates the vehicle path without a prior map.
Autonomous mower field test
The upgraded mower runs the autonomy software while cutting grass outdoors.
Next project
Security patrol route planner
30% less overlap and idle time across patrol shifts.
