Max Robotics — Global Service Robot Directory + US Certification

About

David Millard is a leading researcher at the intersection of robotics, simulation, and machine learning, whose work is fundamentally reshaping how robots learn to interact with the physical world. His primary research areas focus on **differentiable simulation** and **model-based reinforcement learning**, where he develops algorithms that bridge the critical "sim-to-real" gap. Millard’s major contribution is pioneering the use of **interactive differentiable simulators**—a breakthrough that allows robots to not only simulate their environment but also compute precise, gradient-based updates to improve their control policies and physical models directly from data. His landmark 2019 paper, "Interactive Differentiable Simulation" (35 citations), laid the foundation for this paradigm, enabling agents to predict the impact of their actions with unprecedented accuracy. He further advanced the field by integrating neural networks with analytical physics models in "NeuralSim" (13 citations) and by developing methods for **probabilistic inference of simulation parameters** (17 citations) to handle real-world uncertainty. Millard has also tackled the challenging domain of **deformable object manipulation**, using learned recurrent models to track and control fast trajectories of non-rigid materials (10 citations). With a growing citation impact and a reputation for blending rigorous theory with practical robotics, Millard is a rising star whose work is essential reading for anyone interested in creating intelligent, physically-aware autonomous systems.