Robots Now Take Natural Language Commands
Dimensional's OpenClaw agents can now control drones using natural language commands like "Follow the next white car." The system uses the Mavlink protocol and is pip-installable, providing a common control layer for 80% of robots and drones, marking a major step forward in embedded real-time control and human-robot interaction.
The MAVLink protocol, first released in 2009 by Lorenz Meier, has become a de-facto standard for drone communication, with the U.S. government even standardizing it for its small reconnaissance drone architecture. Its lightweight, header-only design is crucial for the resource-constrained embedded systems typical in robotics and UAVs. The protocol's current iteration, MAVLink 2.0, was introduced in early 2017 and offers backward compatibility with the widely adopted version 1.0. OpenClaw itself is a TypeScript-based agentic framework designed for local-first execution and high performance. While it can be run in a lightweight "gateway mode" on a Raspberry Pi 5 with 8GB of RAM, running local AI models for full autonomy requires more substantial hardware, such as a Mac Mini or a mini PC with 16-32GB of RAM. For deeply embedded applications, highly efficient versions like ZeroClaw (written in Rust) and NullClaw (written in Zig) are being developed to run on IoT hardware with only a few megabytes of RAM. The project gained significant attention when its creator, Peter Steinberger, joined OpenAI in February 2026 to work on bringing agentic AI to a wider audience. As part of this transition, OpenClaw was moved into an independent, open-source foundation that continues to be supported by OpenAI, ensuring its continued development and accessibility. This move toward high-level, natural language control is mirrored in the work of major players in the Los Angeles aerospace and defense ecosystem. Northrop Grumman is actively developing "agentic AI" to automatically generate and execute spacecraft commands and mission plans, leveraging NVIDIA's AI platforms for development. This initiative aims to enable more complex missions with faster planning and reduced operational costs. Similarly, Costa Mesa-based Anduril Industries builds its autonomous systems around Lattice, a proprietary AI-powered software platform that serves as a central command and control layer. Lattice fuses data from various sensors to command and control autonomous systems like drones and surveillance towers, demonstrating a parallel industry trend toward integrated, AI-driven mission control. Local space exploration companies are also heavily invested in the underlying technologies. SpaceX's push for automated manufacturing of its Starship and Raptor engines relies on advanced robotics, with plans to integrate AI for increasingly autonomous assembly and off-Earth operations. Job postings for robotics software engineers at Southern California-based Honeybee Robotics, now part of Blue Origin, frequently list C++, Python, and experience with the Robot Operating System (ROS) as key qualifications for developing control systems for space applications.
