OhhO Autonomy

Map it, navigate it, command it in plain language.

• Category: Intelligence
• Accent: violet
• Live app: Open mission control

The robot's autonomy stack — SLAM mapping, Nav2 navigation and a mission planner — driven by a natural-language agent that turns "tidy the kitchen" into a sequence of robot actions.

Overview (hero)

Teleop is for when a human drives; Autonomy is for when the robot drives itself. It maps a space with SLAM, navigates it with Nav2, and runs a mission planner that sequences navigation and manipulation — all orchestrated by a natural-language agent that turns an instruction into a plan.

Highlights

• 2-D & 3-D SLAM mapping
• Nav2 path planning + obstacle avoidance
• Mission planner (navigate → act sequences)
• Natural-language agent (LLM-backed, your choice of model)
• Safe control-mode mux: nav / AI / teleop

What you get

• Perception tells a robot what's around it; Autonomy decides what to do about it. OhhO Autonomy is the behavior layer that turns a powered-on robot into one that moves through the world and completes tasks on its own.
• It builds and localizes against a map with SLAM (2-D and 3-D), plans collision-free paths with Nav2 over a fused costmap, and fuses wheel odometry and IMU through an EKF for reliable pose. A mission planner sequences higher-level jobs — navigate to a named location, then run a manipulation policy — and a control-mode mux arbitrates cleanly between navigation, AI policies and human teleop so they never fight over the wheels.
• On top sits a natural-language agent, backed by the LLM of your choice, that turns 'take the red cup from the kitchen to the bench' into a structured mission: it knows your named locations, can describe what it sees, and asks for clarification when an instruction is ambiguous — then hands the mission to the planner to execute.

Features

• SLAM mapping — Build and localize against 2-D and 3-D maps; switch between mapping and localization modes against a saved map.
• Nav2 navigation — Collision-free path planning and obstacle avoidance over a fused costmap, tuned to the robot's real velocity and acceleration limits.
• Robust localization — An EKF fuses wheel odometry and IMU so pose stays trustworthy even when mecanum wheels slip.
• Mission planner — Sequence navigation and manipulation into a mission — 'go to the kitchen, then pick up the cup' — as a tracked state machine.
• Natural-language agent — An LLM-backed agent maps plain-language instructions to missions, with named locations, scene description and clarification when it's unsure.
• Safe arbitration — A control-mode mux switches between navigation, AI policies and teleop so commands never conflict — and a human can always take over.

How it works

1. Map the space — Drive once while SLAM builds a map, then save it for localization.
2. Name the places — Tag locations — kitchen, dock, bench — the planner and agent can refer to.
3. Give an instruction — Type or speak a task; the agent turns it into a structured mission.
4. Watch it execute — Nav2 and the policies run the mission; the mux keeps teleop override one tap away.

Specs

Plans

Recommended plan: Builder. Builder gives you mapping, navigation and the mission planner on real hardware — the core of autonomy. Add Fleet when you want the natural-language agent and reusable named missions; Forge is for custom behaviors and running the agent on-prem.

FAQ

How is this different from OhhO Pilot? Pilot is for a human operating the robot; Autonomy is for the robot operating itself. They share the same safe control-mode mux, so you can hand control back and forth instantly.

Does the language agent need the cloud? By default it calls a cloud LLM, but the agent layer is optional and model-agnostic — navigation and the mission planner run fully on-robot, and Forge can run the agent on-prem with your own model.

Related products

• OhhO Serve
• OhhO View
• OhhO Pilot