Key Insights

Delivered two coordinated robotics demos combining navigation, voice interaction, and gesture recognition, transforming fragmented PoC repositories into reproducible, deployable showcase systems.

About the Client

A global telecommunications company focused on demonstrating the role of connectivity in enabling real-time, intelligent robotic systems.

The Challenge

The client aimed to build high-impact demos using two robotic platforms (humanoid and quadruped) to showcase coordinated behavior, autonomy, and interaction.

Existing repositories provided a starting point, but were incomplete and unstable. The system required:

• Reliable SLAM navigation and teleoperation
• Low-latency voice interaction and gesture recognition on edge
• Integration across multiple SDKs, sensors, and AI components
• Clear deployment and reproducibility for internal teams

Without consolidation and validation, the demos risked remaining experimental rather than usable showcases.

Marvik’s Approach

We focused on turning fragmented prototypes into working, reproducible systems.

• Robotics Integration & Control: Built and stabilized ROS2-based control stacks, integrated sensors (LiDAR, cameras), and ensured consistent behavior across both robots.
• Voice & Gesture Interaction: Implemented speech-to-text → LLM → text-to-speech pipelines and real-time gesture recognition to trigger robot actions and behaviors.
  
  
• Simulation & Testing: Developed simulation environments (Gazebo) and fallback modes to enable testing without hardware and accelerate iteration.
  
  
• Reproducibility & Deployment: Standardized repositories, created setup scripts, and documented configurations to ensure consistent deployment across environments.

The Results & Impact

• Two functional demos combining navigation, voice, and gesture interaction
• Stable and reproducible systems validated in both simulated and real environments
• Clear deployment guides and documentation for internal teams
• Reduced setup time and improved reliability for live demonstrations

Why This Matters

Robotics demos often fail at the last mile: integration, stability, and reproducibility.

By structuring the system end-to-end and documenting it clearly, this project enabled teams to run, adapt, and extend the demos without re-engineering them from scratch.