I pursue here an idea to overcome the inefficiency of many mechanical hydraulic systems that throttle a constant pressure fluid source to drive hydraulic actuators. The power consumed is that constant pressure times the displacement even when the resistance to the action is minimal. The efficiency of such systems is limited to 1/(peak to average ratio of resistance). “Displacement” here can be taken as either linear displacement, against a force, or volume displacement against a pressure. This was listed as one of the problems to be overcome by the developers of Big Dog by Boston Dynamics. This is why cars have transmissions.

The notion is to have several hydraulic reservoirs at different progressive pressures and open a hydraulic path from each actuator to just one reservoir at a time, to match the changing resistance, including inertial resistance. I can imagine several schemes to manage which reservoir the hydraulic actuator is connected to. For now I imagine a round house like cylinder within a snug housing and a tube terminating in that housing for each pressure reservoir. Turning the cylinder moves an outlet in the cylinder to any one of these sources. The outlet leads to the hydraulic actuator — in effect a single-pole multi throw valve. There are probably better designs.

We must maintain the nominal pressure in each reservoir and in so doing provide energy to the system. To store energy each reservoir might include a gas bladder as a small buffer to keep small volume changes from producing large pressure changes. Pressure fluctuations produce a small Carnot cycle which dissipates some of our energy. A pressure relief valve between neighboring reservoirs avoids excess pressure differences but to rely totally on this obviates the whole idea. A simple check valve in the opposite direction keeps the pressure sequence monotonic. Ideally neither valve would be used.

The raw system energy input goes to pumping fluid from low pressure reservoirs to high pressure ones. I assume here a pump between neighboring reservoirs and a uniform pressure difference. With the gas bladders it is theoretically possible to use a timeshared pump but I doubt that this is feasible. A positive displacement pump is efficient at a constant pressure difference but with varying flow rates. The angular displacement of the shaft is proportional to the volume of the displaced fluid and the pressure difference is proportional to the torque. Consider a tree of mechanical differentials so that the sum of the angular displacements at the pumps is the angular displacement of the ultimate power source. The torques at the pumps in such a system is uniform at one moment and thus the static equilibrium is a uniform set of pressure differences between reservoirs. N.B. Sometimes the energy will be flowing in the reverse direction. There is no theoretical reason to lose this energy! This scheme presumes a rotary power supply of constant torque but variable angular velocity, even negative velocity. I don’t know how best to provide such power.

The above presumes that the software controls the pressure. If it is better to nominally control actuator displacement, then a simple feed back loop between proprioceptive sensors can be arranged, perhaps entirely mechanically. (Like power steering in a car.)

Another alternative scheme should be mentioned. Centrifugal pumps have the approximate property that at constant angular velocity the torque on the shaft, and thus the power are proportional to the flow rate. Further the pressure difference is constant. A set of centrifugal pumps plumbed in series and all turned on a common shaft would probably much simpler than the scheme above. They would provide the same set of hydraulic sources at graduated pressures. I don’t know how close real centrifugal pumps come to this ideal, but I think they do not provide sufficient pressure, even in series, for Big Dog.

I first became aware of this property of centrifugal pumps when I was about 5 years old. When the input or output of a vacuum cleaner is blocked the motor speeds up. Normally when the effect of a motor is thwarted the motor slows down or stops.

2011 Sept 30: Alpha Dog
It occurs to me just now (2012 June 5) that there might be an electric analog to this for controlling systems where energy efficiency is critical, such as in electric vehicles.