![](data:image/svg+xml,<svg display="block" id="1524428453" role="presentation" viewBox="0 0 32 32" xmlns="http://www.w3.org/2000/svg"><path d="M 15.987 0.494 C 16.832 0.005 17.837 -0.127 18.78 0.126 C 19.723 0.379 20.527 0.996 21.014 1.843 C 22.03 3.604 21.427 5.856 19.667 6.875 C 17.441 8.161 13.549 6.583 11 8.056 C 10.96 8.078 10.936 8.121 10.938 8.166 C 10.94 8.212 10.968 8.252 11.01 8.271 C 12.385 8.901 20.21 12.489 20.271 12.505 C 21.861 12.93 22.976 14.36 23.001 16.006 C 23.027 17.652 21.957 19.116 20.381 19.591 C 18.112 20.519 14.511 19.191 12.104 20.582 C 9.554 22.056 8.974 26.219 6.749 27.506 C 5.904 27.994 4.899 28.126 3.956 27.873 C 3.013 27.62 2.21 27.003 1.723 26.157 C 0.707 24.395 1.309 22.143 3.069 21.125 C 5.619 19.651 9.187 21.417 11.736 19.944 C 12.009 19.786 11.955 19.387 11.666 19.262 L 2.727 15.427 C 1.301 15.043 0.242 13.845 0.036 12.383 C -0.17 10.921 0.517 9.477 1.781 8.714 C 1.841 8.674 1.902 8.636 1.965 8.599 C 4.515 7.126 8.083 8.892 10.632 7.418 C 13.182 5.944 13.438 1.968 15.987 0.494 Z" fill="rgb(255, 255, 255)" height="27.99884804175999px" id="vO3ypKHss" transform="translate(4 2)" width="23.00194688367269px"/></svg>)
Observability
See how your machines behave. Capture what matters.
See how your machines behave. Capture what matters.
Physicalfleet monitors your models and devices in production. When confidence drops or behavior looks abnormal, capture camera footage, model traces, and hardware signals—giving your team the context to improve models and applications.
Physicalfleet monitors your models and devices in production. When confidence drops or behavior looks abnormal, capture camera footage, model traces, and hardware signals—giving your team the context to improve models and applications.
metric
task_success_rate
current
71%
expected
95–99%
device
Robot-002
Abnormal behavior
2 min ago
Task success dropped
Robot-002’s task success fell to 71%, below its usual 95–99% range.
Expected range
Recording triggered
14:00 / 14:10
Oct 15
Oct 22
Oct 29
Event recorded
Delivered 1m ago
metric
task_success_rate
current
71%
expected
95–99%
device
Robot-002
Abnormal behavior
2 min ago
Task success dropped
Robot-002’s task success fell to 71%, below its usual 95–99% range.
Expected range
Recording triggered
14:00 / 14:10
Oct 15
Oct 22
Oct 29
Event recorded
Delivered 1m ago
Adaptive baselines, per metric.
Track task success, model confidence, inference latency, and hardware signals. Spot unexpected changes on individual devices and patterns across your fleet—even when the software reports no error.
Event recording
3
/ 4
Robot-002 · VLA v2.5
Captured context
Camera footage
Saved
Model inputs & outputs
Saved
Sensor & motor signals
Saved
Application logs
Saved
Event recording
3
/ 4
Robot-002 · VLA v2.5
Captured context
Camera footage
Saved
Model inputs & outputs
Saved
Sensor & motor signals
Saved
Application logs
Saved
Record the full context.
Capture camera footage, model inputs and outputs, sensor readings, motor responses, and application logs around each event. Review what the machine saw, what the model produced, and what happened next.
Learn from what happens in the field.
Build better models and applications.
Task success · Last 50 runs
Monitoring
Robot-002 · VLA v2.5
Run 10
Run 20
Run 40
Run 60
Run 80
Observe models and machines together.
Follow model confidence, task outcomes, latency, and device health in one platform. View behavior by device, model version, or fleet group.
Capture the moments worth reviewing.
Severity is weighted by how far a deviation sits from baseline, when it's happening, and which metrics are affected. A ranked signal, not a flat list.
Review, compare, and improve.
The right alert reaches the right person every time. Route by severity, by team, or by source — without anyone manually deciding who gets what.
Observe models and machines together.
Follow model confidence, task outcomes, latency, and device health in one platform. View behavior by device, model version, or fleet group.
Task success · Last 50 runs
Monitoring
Robot-002 · VLA v2.5
Run 10
Run 20
Run 40
Run 60
Run 80
Capture the moments worth reviewing.
Severity is weighted by how far a deviation sits from baseline, when it's happening, and which metrics are affected. A ranked signal, not a flat list.
Event recordings · Latest captures
20
Low model confidence
Saved
Robot-002 · Camera + model trace
·
35
Unexpected motor response
Saved
Robot-017 · Motor + sensor signals
·
15
Task failed
Saved
Robot-031 · Camera + application logs
·
Review, compare, and improve.
The right alert reaches the right person every time. Route by severity, by team, or by source — without anyone manually deciding who gets what.
Compare recordings
24 runs
if
task
=
Pick & place
if
fleet_group
=
Warehouse A
and
Include outcomes:
if
outcome
=
Success
or
outcome
=
Failed
and
Model versions
![](data:image/svg+xml,<svg display="block" role="presentation" viewBox="0 0 32 32" xmlns="http://www.w3.org/2000/svg"><g d="M 0 0 L 24 0 L 24 24 L 0 24 Z M 7.858 12.212 L 10.576 13.469 L 10.228 16.592 L 5.369 22.097 C 2.487 20.245 0.533 17.249 0 13.865 Z M 15.11 15.564 L 18.745 21.827 C 16.728 23.245 14.322 24.004 11.856 24 C 11.04 24 10.227 23.918 9.428 23.754 L 10.228 16.591 L 12.286 14.259 Z M 23.462 8.929 C 23.72 9.909 23.858 10.938 23.858 11.999 C 23.858 14.469 23.112 16.764 21.833 18.673 L 15.11 15.564 L 13.533 12.848 L 15.515 10.602 Z M 12.08 0 C 15.56 0.061 18.841 1.633 21.069 4.307 L 15.514 10.602 L 12.587 11.218 L 11.111 8.674 Z M 11.111 8.675 L 10.785 11.596 L 7.859 12.212 L 0.309 8.721 C 1.26 5.386 3.603 2.623 6.739 1.141 Z" fill="transparent" height="24.000015083383285px" id="kqrJ4gA6m" transform="translate(4 4)" width="24px"><path d="M 0 0 L 24 0 L 24 24 L 0 24 Z" fill="transparent" height="24px" id="DuhZHebCq" width="24px"/><path d="M 7.858 0 L 10.576 1.257 L 10.228 4.38 L 5.369 9.885 C 2.487 8.033 0.533 5.037 0 1.653 Z" fill="rgb(21, 227, 255)" height="9.885000000000002px" id="JJEAJseCd" transform="translate(0 12.212)" width="10.576px"/><path d="M 5.682 1.305 L 9.317 7.568 C 7.3 8.986 4.894 9.745 2.428 9.741 C 1.612 9.741 0.799 9.659 0 9.495 L 0.8 2.332 L 2.858 0 Z" fill="rgb(52, 141, 252)" height="9.74101525885936px" id="rQgESjQWF" transform="translate(9.428 14.259)" width="9.316999999999997px"/><path d="M 9.929 0 C 10.187 0.98 10.325 2.009 10.325 3.07 C 10.325 5.54 9.579 7.835 8.3 9.744 L 1.577 6.635 L 0 3.919 L 1.982 1.673 Z" fill="rgb(253, 72, 115)" height="9.744px" id="oy2ZO1kid" transform="translate(13.533 8.929)" width="10.325px"/><path d="M 0.969 0 C 4.449 0.061 7.73 1.633 9.958 4.307 L 4.403 10.602 L 1.476 11.218 L 0 8.674 Z" fill="rgb(255, 199, 0)" height="11.218px" id="Am8Bk4LNT" transform="translate(11.111 0)" width="9.957999999999998px"/><path d="M 10.802 7.534 L 10.476 10.455 L 7.55 11.071 L 0 7.58 C 0.951 4.245 3.294 1.482 6.43 0 Z" fill="rgb(0, 231, 185)" height="11.071px" id="gQ5F0q2E4" transform="translate(0.309 1.141)" width="10.802px"/></g></svg>)
Version A
v2.4
![](data:image/svg+xml,<svg display="block" role="presentation" viewBox="0 0 32 32" xmlns="http://www.w3.org/2000/svg"><g d="M 0 0 L 28 0 L 28 22 L 0 22 Z M 19.811 13.852 C 19.811 10.462 16.973 7.712 13.471 7.712 L 8.19 7.712 L 8.19 13.852 L 13.473 13.852 L 13.473 21.527 L 19.813 21.527 L 19.813 13.852 Z M 0.264 14.363 C 0.264 18.319 3.576 21.526 7.66 21.526 L 11.887 21.526 L 11.887 15.386 L 7.66 15.386 C 7.077 15.386 6.604 14.928 6.604 14.363 L 6.604 13.852 L 0.264 13.852 Z M 13.472 0.038 C 21.35 0.038 27.736 6.223 27.736 13.852 L 27.736 21.526 L 21.396 21.526 L 21.396 13.852 C 21.396 9.614 17.848 6.178 13.472 6.178 L 7.66 6.178 C 7.077 6.178 6.604 6.636 6.604 7.201 L 6.604 7.712 L 0.264 7.712 L 0.264 7.201 C 0.264 3.245 3.576 0.038 7.66 0.038 Z" fill="transparent" height="22px" id="qwMfr9sDS" transform="translate(2 5)" width="28px"><path d="M 0 0 L 28 0 L 28 22 L 0 22 Z" fill="transparent" height="22px" id="SYH5sBkDL" width="28px"/><path d="M 11.621 6.14 C 11.621 2.75 8.783 0 5.281 0 L 0 0 L 0 6.14 L 5.283 6.14 L 5.283 13.815 L 11.623 13.815 L 11.623 6.14 Z" fill="rgb(255, 77, 0)" height="13.815000000000001px" id="JP4pYm_jy" transform="translate(8.19 7.712)" width="11.623px"/><path d="M 0 14.325 C 0 18.281 3.312 21.488 7.396 21.488 L 11.623 21.488 L 11.623 15.348 L 7.396 15.348 C 6.813 15.348 6.34 14.89 6.34 14.325 L 6.34 13.814 L 0 13.814 Z M 13.208 0 C 21.086 0 27.472 6.185 27.472 13.814 L 27.472 21.488 L 21.132 21.488 L 21.132 13.814 C 21.132 9.576 17.584 6.14 13.208 6.14 L 7.396 6.14 C 6.813 6.14 6.34 6.598 6.34 7.163 L 6.34 7.674 L 0 7.674 L 0 7.163 C 0 3.207 3.312 0 7.396 0 Z" fill="rgb(255, 77, 0)" height="21.488000129699707px" id="E93FPLBHy" transform="translate(0.264 0.038)" width="27.471999999999998px"/></g></svg>)
Version B
v2.5
Observe models and machines together.
Follow model confidence, task outcomes, latency, and device health in one platform. View behavior by device, model version, or fleet group.
Task success · Last 50 runs
Monitoring
Robot-002 · VLA v2.5
Run 10
Run 20
Run 40
Run 60
Run 80
Capture the moments worth reviewing.
Severity is weighted by how far a deviation sits from baseline, when it's happening, and which metrics are affected. A ranked signal, not a flat list.
Event recordings · Latest captures
20
Low model confidence
Saved
Robot-002 · Camera + model trace
·
35
Unexpected motor response
Saved
Robot-017 · Motor + sensor signals
·
15
Task failed
Saved
Robot-031 · Camera + application logs
·
Review, compare, and improve.
The right alert reaches the right person every time. Route by severity, by team, or by source — without anyone manually deciding who gets what.
Compare recordings
24 runs
if
task
=
Pick & place
if
fleet_group
=
Warehouse A
and
Include outcomes:
if
outcome
=
Success
or
outcome
=
Failed
and
Model versions
Version A
v2.4
Version B
v2.5
Your intelligence fleet,
finally connected
Connect
© 2026 Physicalfleet. All rights reserved.