Enabling The Next Generation of Autonomous Robotics
Developing Robotic Autonomy is Complicated
Autonomous Robotics has been a field in the making for 40 years, starting with the simplest of machines that had almost no ability to process information. Today, robots are becoming increasingly sophisticated, capable of viewing and understanding the world around them, moving quickly through challenging environments, and automating tasks that most people don't want to do.
However, as these robots become more sophisticated, building and deploying them in the real world become increasingly challenging. These issues stem from various factors, including labor shortages of highly skilled software developers and engineers, inherent challenges of deploying real-world systems in chaotic environments, and the complexities of navigating a challenging regulatory landscape. Coupled with these challenges is the massive investment required in infrastructure to get the robot hardware to customers. Most organizations follow vastly different processes, building their own platforms, hardware, software, and integration tools, resulting in a lack of standardization and requiring specialized knowledge. Even after building a robot, the challenges persist in integrating with customers and gaining their trust, which remains one of the foundational problems in robotics. Consequently, this severely limits the number of organizations and researchers capable of developing these systems, straining small businesses and hindering the ability to deploy robots at scale.
Most of the existing robotics development platforms are an amalgamation of different plugins, technologies, and softwares. Unfortunately, the user interface often resembles something NASA built in the 70s to fly to the Moon, which can be wonderfully useful but baffling to use. Moreover, many platforms still rely on old settings files in XML format, requiring synchronization of assets and offering little intuitive ability to understand the system as a whole, visualize key components of the entire architecture, or provide tools for testing different aspects of robots. Even open-source software like the Robotic Operating System (ROS), while highly useful, can be incredibly difficult to use and lacks deployment tools for robotics developers.
How Can We Simplify This?
At Codex Labs, we firmly believe that these challenges are solvable. In general, building a robot is a simple three-step process: design, build, and deploy. Although each step may involve numerous complexities, the underlying paradigm remains consistent across industries. This led us to ask the question: why does everyone insist on doing everything on their own, and how can we make it easier for companies to build trusted systems?
The Autonomous Robotics Development Lifecycle
To address these issues and propel robotics development forward, we have spent the last 3 years collaborating with experts in robotics, multi-agent systems, physics, and more to chart the next step forward. Our journey began with enhancing the user interface, iterating through hundreds of designs to discern what robotics developers genuinely wanted and needed from their toolset. From there, we delved deep into the core tools to make robotics development easier: simulation, automation, and data analysis. We embarked on integrating these tools with cutting-edge technology, starting with Unreal Engine 5, a leading video game engine. Subsequently, we developed SWARM RDS, an entire robotics firmware that serves as a framework for quickly developing, testing, and simulating various robotic algorithms and solutions. This firmware effortlessly integrates with many different frameworks, including TensorFlow and PyTorch. Moreover, we developed our own Docker-based system for easy deployment and focused on various aspects of the entire development process.
The Need for Advanced User-Friendly Robotics Development Software
One significant challenge we sought to solve was the requirement for powerful computers to run Unreal Engine, which can be expensive and not always accessible. We resolved this challenge by harnessing the power of the cloud and video game streaming, making a photo-realistic environment accessible to anyone from the convenience of a web browser. To further aid developers during testing, we combined this with an interactive User Interface application providing different data outputs.
Recognizing the need for dependable, adaptable, and user-friendly robotics software, we developed the SWARM RDS Firmware. This entire robotics software suite leverages 21st-century software engineering practices, boasting an extendable, high-performance framework. We initiated our offerings with simple systems like quadrotor UAVs and fixed-wing aircraft, with plans to rapidly expand. Inspired by the success of platforms like PX4, we realized the necessity for a higher level of autonomy that is extensible and easy to integrate. This led us to enhance our framework with direct integration to the SWARM RDS Simulation framework, enabling users to efficiently develop, test, and deploy their solutions within our framework. Additionally, we offer support for multi-agent systems, or robot swarms, utilizing cutting-edge technologies to ensure safe and trustworthy robot operations.
A derived benefit of our platform is "Simulation as part of the product." We understand that while many developers can build intriguing robotic platforms, few can effectively sell them to customers without a deep understanding of how to utilize them. To address this challenge, we realized that simulation is part of the solution. Through our SWARM RDS platform, we enable easy simulation setup and access, providing companies and researchers with the ability to swiftly generate interactive prototypes of their platforms. These prototypes can be effortlessly streamed and integrated into their customer-facing applications, instilling trust in the end product.
Our mission as a company is simple: help make building and deploying robots as easy as possible while ensuring that everyone can participate in the next generation of technological innovation.