Two axes, one shared DC link, one point-to-point motion. Nothing about the physics changes as you move the meter or switch the energy model — but the number does. This is the reason energy claims for robot trajectories are not comparable across papers, and it is why a measurement convention has to fix the plane and the model before it fixes anything else.
| Model \ plane | mains | cabinet | DC link | per-axis sum | ratio max/min |
|---|
Each column is one place to put the instrument; compare across a row. Each row is a different definition of "energy" in different units, so the rows are not competing estimates of one quantity and must not be compared to each other — that conflation is itself one of the findings. A published figure is one cell, and nothing requires the author to say which.
Schematic two-axis model: rigid inertia, viscous and Coulomb friction, motor torque constant Kt, winding resistance R, quintic point-to-point profile. Regeneration routing follows the shared-DC-link behaviour disclosed in Fanuc US 8,040,097 (Ebrg = Ebsg − Ebng − RRg) and the chopper-threshold race in Yaskawa JP2005253213A. The ∫|τ| model is the one used in 机器人 43(6):653-663; the I²R dissipation model is Izumi's 散逸エネルギー (JRSJ 15(4), 1997). For intuition, not for mission design — the point is the spread between cells, not any single number.