Researchers unveil Milo, open-source robotic guide dog under $2,000

The introduction of Milo addresses a long-standing accessibility challenge: guide dogs are expensive, in short supply, and require applicants to wait several years. By bringing the cost below $2,000, the robot could democratize mobility assistance for blind and visually impaired people who cannot afford or obtain a traditional guide dog. The developers' choice to build Milo as fully autonomous and independent of cloud computing is significant—it means the robot relies entirely on onboard sensors and AI, making it functional in any environment without requiring internet connectivity or pre-scanned maps. The use of reinforcement learning trained on virtual environments in a custom simulator suggests the developers invested in robust generalization, teaching Milo to handle diverse real-world scenarios it may not have explicitly encountered during training.

The decision to release Milo as open-source—including hardware designs, simulation software, and pre-trained models—is a deliberate move to lower barriers to further development and iteration. This approach invites researchers and engineers outside the original team to contribute improvements, test variations, and adapt the design for different contexts. The evaluation on both indoor and outdoor obstacle courses indicates the developers tested the system under realistic conditions, though the article does not report specific accuracy or speed metrics from those tests.