R&D & Engineering
Selling the robot is where our work starts. Our real job is building the software and integration layer that turns a robot into a working mission: navigation, teleoperation, embedded autonomy and operations software.
Six Field-Validated Engineering Tracks
Autonomous Navigation
LiDAR-inertial odometry (FAST-LIO class), Nav2-based path planning, relocalization and map management. An end-to-end navigation stack built and field-validated on a wheeled quadruped: mapping, route following, obstacle avoidance and dock-based localization.
Teleoperation & Remote Command
Low-latency WebRTC video with browser-based driving; humanoid teleoperation via XR headsets. Multi-layer safety: an ARM/DISARM-disciplined command path, heartbeat watchdog, forward-obstacle protection and software e-stop.
Embedded Autonomy
An architecture where navigation and mission execution run entirely on the onboard NVIDIA Jetson, with no external PC in the loop. Watchdog supervisor, per-service auto-restart and self-check at boot.
RL Locomotion & Simulation
Isaac- and MuJoCo-based reinforcement learning pipelines, sim-to-real transfer and locomotion/balance policy development on the Unitree RL Gym ecosystem. Open to academic collaborations and customer POCs.
Fleet & Operations Software
A browser-based command panel: live telemetry, camera streams, map and mission planning, patrol routes, event logs and reporting. The interface layer that lets an operations team run robots like shift equipment.
SDK / API & Systems Integration
Low-level control via unitree_sdk2 and ROS 2, DDS networking, multi-host communication and enterprise integrations (access control, IoT, reporting). Mission APIs and floor-flow integration on the Pudu fleet.
From Survey to 24/7 Operations, Against Measurable Acceptance Criteria
Site, mission and acceptance criteria
Platform, sensor and software selection
End-to-end validation on a controlled site
Safety layers + operator training
24/7 monitoring, maintenance and iteration
Our safety principles apply at every stage: staged commissioning, speed limits, watchdog supervision and an e-stop that works in every scenario.
Our Focus Areas in the Open-Source Ecosystem
We actively develop and evaluate within the open-source robotics ecosystem around ROS 2, Nav2 and modern RL frameworks. In enterprise deployments we validate, harden and adapt these building blocks to your facility — and we'll build a solution of the same depth for you.
| Focus Area | How We Use It |
|---|---|
| Official Unitree toolchain (unitree_sdk2, unitree_ros2) | The base layer of all our ROS 2 integrations |
| RL locomotion training (Isaac / MuJoCo pipelines) | Core of our locomotion policy training pipeline; sim-to-real transfer |
| Quadruped autonomous navigation | Our architecture benchmarking and field validation work |
| Learning-based navigation (VLN-class models) | Evaluation of deployment approaches |
| Wheeled platform driver layer (Go2-W class) | Our low-level control work on hybrid platforms |
| Autonomous carrier scenarios | Evaluated in logistics mission designs |
| Deep-RL locomotion policies | Our policy architecture research |
| Constraint-aware navigation | Our keep-out zone and virtual wall approaches |
Talk to the team behind these deployments.
Let our enterprise sales and engineering team plan your solution.
Deploy Autonomy That Works at Your Facility
Describe your mission; get a clear engineering proposal with survey, architecture and acceptance criteria.