DVIDIA · PiPER retention v0.2
Checking grip retention without changing the squeeze.
Three authored hold strategies across twelve fresh synthetic scenes. The primary question is whether an admitted grasp remains held continuously for five simulated seconds without a false confirmation. A second reported outcome also requires the TCP to stay within the same five-second pose bounds for every variant.
Simulation-only outcomes. No learned policy, transport, physical sensor validation or hardware qualification.
Authored synthetic PiPER probe and short-pause controls with privileged simulator state. Fresh finite development cells; no learning, full placement, physical sensing or hardware qualification.
Measured Joint
12 scheduled · 12/12 valid · 4 first admissions · 0 false initial admissions · 4/12 with complete five-second follow-up · 2 native grip holds · 2 grip plus stationary-tool pauses · 0 false confirmations · 0 conservative abstentions
| Attempt | Scene | Variant | Valid run | First admission | Native prefix qualified | Native five-second grip hold | Grip + stationary tool | Controller retained | False confirmation | Conservative abstention | Step p95 | Primary reason / error |
|---|
| m45-f75-p0-measured_joint | m45-f75-p0 | measured_joint | yes | 11.740 s · probe 0 | yes | yes | yes | yes | no | no | 0.003 s | success |
| m45-f75-p1-measured_joint | m45-f75-p1 | measured_joint | yes | 12.040 s · probe 0 | yes | yes | yes | yes | no | no | 0.003 s | success |
| m45-f50-p0-measured_joint | m45-f50-p0 | measured_joint | yes | none | — | incomplete follow-up | incomplete follow-up | not evaluated | — | — | 0.003 s | no_admission |
| m45-f50-p1-measured_joint | m45-f50-p1 | measured_joint | yes | none | — | incomplete follow-up | incomplete follow-up | not evaluated | — | — | 0.002 s | no_admission |
| m45-f20-p0-measured_joint | m45-f20-p0 | measured_joint | yes | none | — | incomplete follow-up | incomplete follow-up | not evaluated | — | — | 0.003 s | no_admission |
| m45-f20-p1-measured_joint | m45-f20-p1 | measured_joint | yes | none | — | incomplete follow-up | incomplete follow-up | not evaluated | — | — | 0.004 s | no_admission |
| m55-f75-p0-measured_joint | m55-f75-p0 | measured_joint | yes | 11.760 s · probe 0 | yes | no | no | no | no | no | 0.003 s | native_predicate_lost |
| m55-f75-p1-measured_joint | m55-f75-p1 | measured_joint | yes | 12.060 s · probe 0 | yes | no | no | no | no | no | 0.003 s | native_predicate_lost |
| m55-f50-p0-measured_joint | m55-f50-p0 | measured_joint | yes | none | — | incomplete follow-up | incomplete follow-up | not evaluated | — | — | 0.003 s | no_admission |
| m55-f50-p1-measured_joint | m55-f50-p1 | measured_joint | yes | none | — | incomplete follow-up | incomplete follow-up | not evaluated | — | — | 0.003 s | no_admission |
| m55-f20-p0-measured_joint | m55-f20-p0 | measured_joint | yes | none | — | incomplete follow-up | incomplete follow-up | not evaluated | — | — | 0.003 s | no_admission |
| m55-f20-p1-measured_joint | m55-f20-p1 | measured_joint | yes | none | — | incomplete follow-up | incomplete follow-up | not evaluated | — | — | 0.003 s | no_admission |
Issued Joint
12 scheduled · 12/12 valid · 4 first admissions · 0 false initial admissions · 4/12 with complete five-second follow-up · 4 native grip holds · 0 grip plus stationary-tool pauses · 0 false confirmations · 0 conservative abstentions
| Attempt | Scene | Variant | Valid run | First admission | Native prefix qualified | Native five-second grip hold | Grip + stationary tool | Controller retained | False confirmation | Conservative abstention | Step p95 | Primary reason / error |
|---|
| m45-f75-p0-issued_joint | m45-f75-p0 | issued_joint | yes | 11.740 s · probe 0 | yes | yes | no | yes | no | no | 0.002 s | success |
| m45-f75-p1-issued_joint | m45-f75-p1 | issued_joint | yes | 12.040 s · probe 0 | yes | yes | no | yes | no | no | 0.002 s | success |
| m45-f50-p0-issued_joint | m45-f50-p0 | issued_joint | yes | none | — | incomplete follow-up | incomplete follow-up | not evaluated | — | — | 0.004 s | no_admission |
| m45-f50-p1-issued_joint | m45-f50-p1 | issued_joint | yes | none | — | incomplete follow-up | incomplete follow-up | not evaluated | — | — | 0.002 s | no_admission |
| m45-f20-p0-issued_joint | m45-f20-p0 | issued_joint | yes | none | — | incomplete follow-up | incomplete follow-up | not evaluated | — | — | 0.003 s | no_admission |
| m45-f20-p1-issued_joint | m45-f20-p1 | issued_joint | yes | none | — | incomplete follow-up | incomplete follow-up | not evaluated | — | — | 0.004 s | no_admission |
| m55-f75-p0-issued_joint | m55-f75-p0 | issued_joint | yes | 11.760 s · probe 0 | yes | yes | no | yes | no | no | 0.003 s | success |
| m55-f75-p1-issued_joint | m55-f75-p1 | issued_joint | yes | 12.060 s · probe 0 | yes | yes | no | yes | no | no | 0.003 s | success |
| m55-f50-p0-issued_joint | m55-f50-p0 | issued_joint | yes | none | — | incomplete follow-up | incomplete follow-up | not evaluated | — | — | 0.004 s | no_admission |
| m55-f50-p1-issued_joint | m55-f50-p1 | issued_joint | yes | none | — | incomplete follow-up | incomplete follow-up | not evaluated | — | — | 0.003 s | no_admission |
| m55-f20-p0-issued_joint | m55-f20-p0 | issued_joint | yes | none | — | incomplete follow-up | incomplete follow-up | not evaluated | — | — | 0.003 s | no_admission |
| m55-f20-p1-issued_joint | m55-f20-p1 | issued_joint | yes | none | — | incomplete follow-up | incomplete follow-up | not evaluated | — | — | 0.004 s | no_admission |
Cartesian
12 scheduled · 12/12 valid · 4 first admissions · 0 false initial admissions · 4/12 with complete five-second follow-up · 4 native grip holds · 0 grip plus stationary-tool pauses · 0 false confirmations · 4 conservative abstentions
| Attempt | Scene | Variant | Valid run | First admission | Native prefix qualified | Native five-second grip hold | Grip + stationary tool | Controller retained | False confirmation | Conservative abstention | Step p95 | Primary reason / error |
|---|
| m45-f75-p0-cartesian | m45-f75-p0 | cartesian | yes | 11.740 s · probe 0 | yes | yes | no | no | no | yes | 0.002 s | success |
| m45-f75-p1-cartesian | m45-f75-p1 | cartesian | yes | 12.040 s · probe 0 | yes | yes | no | no | no | yes | 0.004 s | success |
| m45-f50-p0-cartesian | m45-f50-p0 | cartesian | yes | none | — | incomplete follow-up | incomplete follow-up | not evaluated | — | — | 0.003 s | no_admission |
| m45-f50-p1-cartesian | m45-f50-p1 | cartesian | yes | none | — | incomplete follow-up | incomplete follow-up | not evaluated | — | — | 0.003 s | no_admission |
| m45-f20-p0-cartesian | m45-f20-p0 | cartesian | yes | none | — | incomplete follow-up | incomplete follow-up | not evaluated | — | — | 0.003 s | no_admission |
| m45-f20-p1-cartesian | m45-f20-p1 | cartesian | yes | none | — | incomplete follow-up | incomplete follow-up | not evaluated | — | — | 0.003 s | no_admission |
| m55-f75-p0-cartesian | m55-f75-p0 | cartesian | yes | 11.760 s · probe 0 | yes | yes | no | no | no | yes | 0.003 s | success |
| m55-f75-p1-cartesian | m55-f75-p1 | cartesian | yes | 12.060 s · probe 0 | yes | yes | no | no | no | yes | 0.004 s | success |
| m55-f50-p0-cartesian | m55-f50-p0 | cartesian | yes | none | — | incomplete follow-up | incomplete follow-up | not evaluated | — | — | 0.003 s | no_admission |
| m55-f50-p1-cartesian | m55-f50-p1 | cartesian | yes | none | — | incomplete follow-up | incomplete follow-up | not evaluated | — | — | 0.003 s | no_admission |
| m55-f20-p0-cartesian | m55-f20-p0 | cartesian | yes | none | — | incomplete follow-up | incomplete follow-up | not evaluated | — | — | 0.004 s | no_admission |
| m55-f20-p1-cartesian | m55-f20-p1 | cartesian | yes | none | — | incomplete follow-up | incomplete follow-up | not evaluated | — | — | 0.003 s | no_admission |
How to read the outcomes
“Native five-second grip hold” is the simulator outcome after first admission. “Grip + stationary tool” additionally requires the TCP to remain within 4 mm and 0.05 rad of its admission pose throughout the five-second window. “Controller retained” is the controller's corresponding qualification claim. A false confirmation is counted only when complete follow-up exists, the native grip hold fails, and the controller claims retention. Conservative abstention is shown separately when the native hold succeeds and the controller does not claim retention. Initial admission against the native prefix is a separate event. Incomplete follow-up is not counted as either mismatch.
Step p95 is unpaced full control-step compute wall time (including dense telemetry); it is not sensor latency or a real-time guarantee. Runs use privileged simulator state and do not establish physical retention.