Abstract
Web-based applications are highly accessible to users, providing rich, interactive content while eliminating the need to install software locally. However, evolutionary robotics (ER) has faced challenges in this domain as web-based technologies have not traditionally been amenable to 3D physics simulations. Traditionally, physics-based simulations require a local installation and a high degree of user knowledge to configure an environment. The emergence of Javascript-based physics engines now enables complex simulations to be executed directly in web browsers. These developments create opportunities for ER research to reach new audiences by increasing accessibility. In this work, we introduce two web-based tools built to facilitate the exchange of ideas with other researchers as well as outreach to K-12 students and the general public. The first tool is intended to distribute and exchange ER research results, while the second is a completely browser-based implementation of an ER environment.
Introduction
Evolutionary Robotics (ER) involves the use of evolutionary algorithms to design robot morphologies and controllers. This process is computationally intensive and has historically relied on "heavy" 3D physics engines such as ODE, Bullet, or PhysX. These engines typically require local installation, specific hardware configurations, and significant technical expertise to set up. Furthermore, sharing results has often been limited to static video formats, which restrict the viewer to the author's chosen camera angles and playback speeds. This lack of interactivity hinders the exchange of ideas and limits the potential for public engagement.
By leveraging WebGL and Javascript-based physics engines, we can bring ER out of specialized labs and into the browser. This shift significantly lowers the "activation energy" required for researchers and students to engage with the field.
The WebGL-Based Visualizer
The first tool developed is a WebGL-based visualizer designed to decouple physics simulation from rendering. This tool utilizes a standardized log file format that contains the initial simulation configuration and the state of all bodies at each timestep.
By using WebGL and the ThreeJS library, the visualizer can render these logs in any modern browser. This allows users to interactively rotate the camera, change body colors, and adjust playback speeds. This approach allows researchers using any physics engine (local or web-based) to share their results in a format that is far more informative than a standard video.
Evolve-a-Robot: Browser-Based Evolution
The second tool, "Evolve-a-Robot," is a full-fledged, browser-based ER environment. It utilizes PhysiJS—a Javascript port of the Bullet physics engine—to handle real-time 3D physics simulations directly in the browser.
The evolution in this environment is driven by a 1+1 evolutionary algorithm, a form of hill-climbing where a parent and a single mutated offspring compete. To provide transparency into the evolutionary process, the tool integrates D3.js to generate dynamic, real-time scatter plots of fitness levels as the evolution progresses. This demonstrates that not only the results of ER, but the process itself, can be made accessible and monitorable in real-time.
Technical Implementation and Capabilities
The system demonstrates the maturity of Javascript and WebGL, showing they are capable of handling the rigors of 3D rigid-body dynamics. To showcase the system's flexibility, we utilized three distinct "animat" (artificial animal) models:
- A quadruped.
- An octopod.
- A Karl Sims-inspired animat based on seminal work in evolving 3D morphology.
The primary insight of this work is that accessibility is a core feature. Making ER "one click away" transforms it from a niche academic pursuit into a powerful tool for K-12 outreach and rapid cross-institutional collaboration.
Limitations and Future Directions
While these tools represent a significant step forward, several limitations remain. In the current prototype of "Evolve-a-Robot," evolutionary parameters such as mutation rates are fixed. Future iterations will allow these to be user-definable to encourage deeper experimentation. Additionally, while browser-based physics is advancing, it still lags behind native C++ implementations for massive-scale simulations.
A major future goal is the implementation of a distributed Genetic Algorithm (GA). In this proposed setup, a central server would distribute genomes to many different users' browsers. The browsers would perform the physics evaluations locally and return the results to the server, effectively creating a volunteer-computing grid for robotics research.
Summary updated