Teaching robots to dodge balls thrown at them is easier said than done. There is a lot goes into training robots to detect oncoming balls to adjust. Humanoid Dodgeball explores that. It is a reinforcement learning framework that enables robots to avoid obstacles with agility. It enables robots to take evasive maneuvers against incoming projectiles. Researchers from National University of Singapore and EngineAI achieved this
More like this ➡️ here
through threat-aware reward shaping, while multiple direction-specific adversarial motion priors regularize the learned behaviors toward stable and diverse human-like motions.
These folks designed rewards based on the predictive and near-field threats to encourage the humanoid to dodge dynamic obstacles.
[HT]
*Our articles may contain aff links. As an Amazon Associate we earn from qualifying purchases. Please read our disclaimer on how we fund this site.
