Max Robotics — Global Service Robot Directory + US Certification

About

Dhruva Tirumala is a leading researcher at the intersection of robotics and artificial intelligence, whose work is redefining what is possible with deep reinforcement learning (deep RL) for legged locomotion and complex motor skills. His most celebrated contribution is pioneering the use of deep RL to teach a low-cost, miniature bipedal robot to play a simplified one-versus-one game of soccer—a feat that required synthesizing agile, safe, and sophisticated full-body movements. This landmark 2024 paper, with 147 citations, demonstrates that complex behavioral strategies can be composed from learned skills in dynamic environments. To address the data efficiency challenges inherent in such tasks, Tirumala developed Hindsight Off-policy Options (HO2), a novel algorithm that robustly trains hierarchical policies end-to-end from limited data. He has further advanced the field by proposing hierarchical latent mixture models for learning transferable motor skills and by demonstrating end-to-end robot soccer policies using only egocentric RGB vision and onboard computation. Through this body of work, Tirumala is pushing the boundaries of how robots can learn, adapt, and perform in the real world.