Subsystem · SW

Autonomy & software

Localization, sensor fusion, and traversal planning — architecting the full autonomy stack.

Lead filledTaking members

The rover has to cross rough, unmapped natural terrain without GPS — building its own understanding of the ground beneath it in real time. There is no prior map to localize against and no absolute position fix to correct drift.

What makes it hard

Every part of the stack degrades the others. Odometry drifts, so the map warps; the map warps, so the planner commits to paths that no longer exist; the planner commits, so the vehicle ends up somewhere the estimator did not predict. Getting this to hold together over a long run is the central technical problem of the team.

What you’d get out of it

The autonomy stack in full, not a slice of it — and the chance to work on state estimation and planning problems that most people do not touch until graduate school or industry.

What you'd work on
Localization without GPS — visual/inertial odometry and drift management
Sensor fusion across cameras, IMU, and depth
Terrain understanding: what is traversable, and at what cost
Path and motion planning over a map the rover builds as it goes
The software architecture, tooling and simulation everyone else develops against

Tools

ROS 2 C++, Python Open3D / PCL, OpenCV Gazebo / Isaac Sim

Member openings

Join the subsystem and own a piece of it. No lead experience needed.

Perception engineer

Turn raw camera and depth data into a terrain representation the planner can trust.

OPEN

State estimation engineer

Localization and sensor fusion with no GPS to fall back on.

OPEN

Planning engineer

Traversability costing and path planning over a map that is being built in real time.

OPEN

Simulation & infrastructure

The sim environment, logging, and replay tooling that makes everything else debuggable.

OPEN
Apply to Autonomy & software