גנגליה

Gangelia develops a solution for AI and robotics teams that combines recorded data with simulation and physics-based augmentation. It generates failure and recovery scenarios that are difficult and costly to collect in the real world, helping reduce physical testing and improve robot reliability. Target users are robot manufacturers, robotics application developers and research labs. Key applications include training robotic arms and humanoids for pick-and-place tasks, and testing resilience to environmental changes and sensor faults. The solution targets vision-language-action (VLA) models, robotic policies, and failure detection and recovery models. Augmentation introduces physical variations and sensor and action disturbances, with labels identifying failures and recovery outcomes. Inputs include task recordings, sensor data and robot states. Outputs include labeled training and testing datasets, adapted models and performance metrics.