Skip to main content
Core autonomous behavior on MARS is split across navigation, manipulation, and interaction. MARS supports multiple navigation modes through the Innate Controller App.

Navigation

Uses a pre-built map and LiDAR localization so MARS knows where it is and can place memories spatially.

Mapfree

Avoids obstacles using onboard sensing only, without pre-built map localization or mapping.

Mapping

Dedicated mode to create and save a new map for later navigation.
Navigation mode runs on a saved map and localizes with LiDAR. It is the mode to use when you want reliable repeatable paths and spatial memory tied to places in the environment.

​
Mapfree mode

Mapfree mode uses the local costmap from onboard sensors and does not require a pre-built map. It is best for fast deployment in new environments, dynamic spaces, simple point-to-point tasks, and outdoor scenarios.

​
Mapping mode

1

Open the Innate Controller App

Start from the app home screen.
2

Go to Configuration -> Mapping

Open the mapping configuration screen.
3

Select Create New Map

Start a fresh map for the current environment.
4

Drive through the environment

Move MARS across rooms, transitions, and key landmarks.
5

Save the map

Save once coverage looks complete.
During mapping, keep the arm in a safe resting position, drive slowly for cleaner map quality, and cover corners and doorways.

​
For developers

Navigation mode is published on /nav/current_mode with values mapfree, mapping, and navigation.

ROS2 Topics

Live streams including /nav/current_mode, LiDAR, and state telemetry.

ROS2 Services

Request/response interfaces for mode and map operations.

ROS2 Actions

Long-running goals with feedback and cancellation.

​
Manipulation

MARS supports two manipulation methods:
  1. Standard CV + code-defined skills: use explicit logic and perception pipelines for deterministic task behavior.
  2. End-to-end AI policies: train manipulation skills from demonstrations and deploy the trained policy on robot.

Code + CV Manipulation

Build manipulation behaviors with code-defined skills and robot interfaces.

End-to-End AI Manipulation

Use foundation models for mobile manipulation with MARS.

​
Interaction

MARS can interact through movement, listening, and speaking.

Head Movement

Move and orient the head for active sensing and expressive behavior.

Listening

Receive voice and audio context through onboard sensing and input channels.

Speaking

Speak responses and status naturally while BASIC is running.