Boston Dynamics designed new hydraulic control valves with novel circuit designs that performed computing functions integrated into the valve itself, making hydraulic systems much more efficient, smaller, and lighter than previously possible.

Boston Dynamics developed an integrated hydraulic power supply about the size of a football weighing 5 kilograms that produces 5 kilowatts of power and incorporates motor, pump, filters, heat exchanger, and valves in a single compact package.

Hardware innovation remains important and necessary for robotics; people who believe hardware innovation is no longer needed are mistaken.

Compliance (spring-like elasticity) in robot leg design is important to how natural walking and running works, with proper spring behavior enabling efficient motion transfer and stability.

In 3D flips with a 2-legged robot, the mathematical and algorithmic calculation is fundamentally the same as planar flips, but must maintain balance in additional degrees of freedom beyond the primary somersault plane.

Boston Dynamics achieved the first 3D somersault with a 2-legged robot while at MIT, with graduate student Russ Tedrake who was also a champion gymnast before becoming an engineer.

WildCat was a quadruped robot that achieved speeds of 19 mph on flat terrain and was likely the fastest quadruped robot ever built.

For acrobatic maneuvers like somersaults where the robot must stick a landing, the control system must plan much further ahead than normal locomotion, pre-calculating momentum and rotation requirements at launch to ensure the landing is within reach.

For typical robot locomotion like walking on even terrain, a limited prediction horizon of approximately 1-1.5 seconds is sufficient, with the control system recalculating at roughly 10 times per second, updating predictions based on current obstacles and foot placement.

Sticking a landing after a flip is very difficult and is widely underrated as an engineering challenge because once in the air, the robot has much less control authority over what happens.

Boston Dynamics' earliest flipping experiments were performed with a planar robot (moving on a 2D surface restricted to a spherical boom) where the landing calculation was relatively simple: determining jump height to provide rotation time and ensuring feet land in correct positions with momentum dissipation.

Russ Tedrake's work on stabilizing complicated robot maneuvers was made possible by his background as a champion gymnast before becoming an engineer, allowing him to translate physical intuition about how humans perform acrobatic maneuvers into mathematical algorithms.

Humans, including most gymnasts and athletes, do not actually understand the physics of how they perform their movements; they learn through coaching and practice but typically lack explicit physical and mathematical understanding of the mechanics.

Current Boston Dynamics robots still do not walk as gracefully or naturally as humans, even though running performance is close to human speed and seems more natural than walking.

Boston Dynamics' early organizational motto was 'you have to run before you can walk,' reflecting the strategy of developing running robots before perfecting walking robot technology.

To achieve sufficient rotation rate in 3D somersaults, robots must perform tucking motions (withdrawing the legs) to accelerate rotation, similar to how human gymnasts tuck to spin faster.

WildCat robot, powered by a small racing go-kart engine, was extremely loud, generating complaints from people three buildings away.

The basic design of hydraulic control valves had remained largely unchanged since the 1950s when they were designed for airplane safety, with little meaningful innovation despite potential for significant improvement through new circuit designs and integrated components.

Dynamic control approaches that predict the future motion of a robot over several seconds ahead and pre-emptively adjust control signals produce more natural movement than reactive servo approaches that only correct based on the robot's current state.

Building robots that perform human-like movements contributes to scientific understanding of how humans actually move, since in the process of formalizing human movement as robot algorithms, engineers discover physical principles humans use implicitly.