Skip to content
See the work
All work
Autonomous Navigation on ROSMASTER: cover image

Navigation

Starting

Autonomous Navigation on ROSMASTER

SLAM and navigation on a mobile rover with Jetson Orin NX onboard, starting from measured bring-up instead of assumed performance.

View the code on GitHub
How the system fits together
  1. PhysicalOnboard sensors
  2. Edge computeOrin NX bring-up baseline
  3. Edge computeSLAM
  4. Edge computePath planning
  5. PhysicalMotor control

The problem

Mobile robots fail differently from arms. Localization drifts, maps go stale, and compute is shared among perception, planning, and control on one small board.

Approach

  • Establish a day-one bring-up baseline on the Orin NX before adding workloads.
  • Build SLAM and navigation on ROS 2, measuring the compute cost of each layer as it is added.

Status

Starting. Bring-up baseline is captured. The first SLAM session was recorded on the Orin with the rover stationary (2026-09-16): the vendor SLAM Toolbox stack’s resource cost, about 11 W board input power at the median and a 647 MB peak for the mapping process. On 2026-10-30 that recording was replayed into a fresh mapping node on the rover’s own board, which built a small map. No driven session has run yet, so loop closure, drift and re-localisation are still to be measured.