MIT associate professor Jack Manchester discusses how simple optimization, differentiable simulation, and compact control methods can support robotics across aquatic, terrestrial, and space applications, at a VinMotion seminar.
Manchester frames planning, control, perception, and estimation as optimization problems. He connects model predictive control, reinforcement learning, and diffusion policies as related approaches, and describes TinyMPC running at 200 Hz on a small drone and RLU-QP reaching 2.5 kHz for simulated Atlas whole-body control.
Manchester shows how least-action mechanics and constraints can model impacts, friction, and contact without smoothing away discontinuities. His group used this approach to build a differentiable fluid and robot simulator, then optimized an eight-link eel robot’s swimming gait and rapid turning maneuver. The hardware followed simulated joint commands closely, even in largely open-loop trials.
For satellite formation navigation, the group combined inter-satellite radio ranges with an anchor measurement and estimated uncertain atmospheric drag from about an hour of on-orbit data. Manchester also describes calibrating a deployable structure simulator from camera tracking, using the model to infer joint behavior and forces that were not directly measured.
Manchester presents predictive sampling that tests many candidate actions in parallel, along with a spline representation that helped it run on a quadruped at roughly 30 to 40 Hz. For heavier object manipulation, a learned walking controller reduced the MPC search from about 30 dimensions to eight and allowed updates at 5 Hz. He also reports that trajectory optimization still takes substantial tuning, and argues that linear models can provide useful control and insight.
Briefs summarise talks and interviews; timestamps open the video at that moment. Claims in a talk are the speaker's own.
A Boston Dynamics Prototype Operations team member explains how the company builds and improves Atlas, from precision machining through assembly and testing, alongside its engineers.
Boston Dynamics engineers discuss the design, simulation, and dexterity of Atlas’s new hands.
Boston Dynamics’ robotics team discusses Atlas’s path to industrial work at Hyundai Motor Group’s Robotics Metaplant Application Center, with a host visiting the facility.
Seth Davis, senior program manager on Boston Dynamics’ Atlas team, describes Atlas’s World Cup halftime role and how the robot is being developed for factory work.
The Atlas Director of Robot Behavior at Boston Dynamics talks with NEXT host John Koetsier about Atlas’s design, robot intelligence, factory integration, and why legs can suit industrial work.
John Koetsier of NEXT speaks with Alexandra Sukharev, a vice president at Bessemer Venture Partners, about which companies may succeed in humanoid robotics and the challenges of real world deployment.
