Compute Regeneration through Re-Productization: Reorganizing Retired Edge AI Silicon into a Delivered Multilingual Service Robot
Compute Regeneration through Re-Productization: Reorganizing Retired Edge AI Silicon into a Delivered Multilingual Service Robot
This research paper introduces GEdgeBot, a multilingual service robot constructed around a retired 32 GB Jetson AGX Xavier. The study details the system architecture where AI and orchestration run on the Xavier while motor control is handled by an STM32F407. Historical measurements report 1,782 MB GPU memory usage with specific models resident, noting a direct-control p99 deviation rising to 49.93 ms under concurrent AI load. The authors conduct audits of the six-mode control state machine and a 14-scenario replay, concluding that results support an existence claim and reproducible architectural hypothesis rather than fleet-level reliability.
The authors present a case study of GEdgeBot, a service robot utilizing retired Jetson AGX Xavier hardware to run multilingual AI services, reporting specific latency metrics and control state machine audits while explicitly noting that results support an existence claim rather than fleet-level reliability.
Source: arXiv Robotics — research abstracts · Read original article ↗
Loading article text…
What the source reports
Publisher-reported claims, with original evidence. These results have not been independently verified by RoboSignal.
Reported numbers
GPU memory
1,782
View original evidence
type measurements report 1,782 MB GPU memory with YOLO, Whisper-small, and NLLB-200 resident, a direct-control p99 deviation rising from 0.24 to 49.93 ms under concurrent AI load, and six of six trigger-to-follow demonstrations at 4.35 s mean latency, with conversational voice unavailable. A November campaign restores an isolated inference stack, re-measures AI workloads, and repeats control-frame runs. Driving two e
Open source S4direct-control p99 deviation
0.24–49.93
View original evidence
type measurements report 1,782 MB GPU memory with YOLO, Whisper-small, and NLLB-200 resident, a direct-control p99 deviation rising from 0.24 to 49.93 ms under concurrent AI load, and six of six trigger-to-follow demonstrations at 4.35 s mean latency, with conversational voice unavailable. A November campaign restores an isolated inference stack, re-measures AI workloads, and repeats control-frame runs. Driving two e
Open source S4trigger-to-follow demonstration latency
4.35
View original evidence
type measurements report 1,782 MB GPU memory with YOLO, Whisper-small, and NLLB-200 resident, a direct-control p99 deviation rising from 0.24 to 49.93 ms under concurrent AI load, and six of six trigger-to-follow demonstrations at 4.35 s mean latency, with conversational voice unavailable. A November campaign restores an isolated inference stack, re-measures AI workloads, and repeats control-frame runs. Driving two e
Open source S4frame interval
50
View original evidence
xecution organisations with the same 6,000 measured 50 ms frame intervals shows the finite-pulse boundary removes host jitter from the pulse train but not the command start phase; its stopping bound is host-enforced, while the MCU 16-bit pulse field is unclamped. A six-check static audit exposes a host--MCU queue mismatch, and a 14-scenario replay executes the audited main() dispatcher across six modes with zero phys
Open source S5contract-checked frames
54
View original evidence
ical device opens and 54 contract-checked frames. Results support an existence claim and a reproducible architectural hypothesis, not fleet-level reliability. The 208 direct-control ratio remains unreplicated; MCU waveform capture and on-site fault injection remain open.
Open source S6control modes
6
View original evidence
xecution organisations with the same 6,000 measured 50 ms frame intervals shows the finite-pulse boundary removes host jitter from the pulse train but not the command start phase; its stopping bound is host-enforced, while the MCU 16-bit pulse field is unclamped. A six-check static audit exposes a host--MCU queue mismatch, and a 14-scenario replay executes the audited main() dispatcher across six modes with zero phys
Open source S5
- paper: not_reported
Source excerpts and review record
Automatically extracted; no manual editorial approval recorded.
type measurements report 1,782 MB GPU memory with YOLO, Whisper-small, and NLLB-200 resident, a direct-control p99 deviation rising from 0.24 to 49.93 ms under concurrent AI load, and six of six trigger-to-follow demonstrations at 4.35 s mean latency, with conversational voice unavailable. A November campaign restores an isolated inference stack, re-measures AI workloads, and repeats control-frame runs. Driving two e
Open source S4
xecution organisations with the same 6,000 measured 50 ms frame intervals shows the finite-pulse boundary removes host jitter from the pulse train but not the command start phase; its stopping bound is host-enforced, while the MCU 16-bit pulse field is unclamped. A six-check static audit exposes a host--MCU queue mismatch, and a 14-scenario replay executes the audited main() dispatcher across six modes with zero phys
Open source S5
ical device opens and 54 contract-checked frames. Results support an existence claim and a reproducible architectural hypothesis, not fleet-level reliability. The 208 direct-control ratio remains unreplicated; MCU waveform capture and on-site fault injection remain open.
Open source S6
arXiv:2611.11698v1 Announce Type: new Abstract: Useful edge hardware can lose product support while retaining substantial computation. We treat this as systems design: compute regeneration here means re-productization, retaining the compute module while redesigning the surrounding product organization. Our case is GEdgeBot, a multilingual service robot built around a retired 32 GB Jetson AGX Xavier. A source-grounded
Open source S2
Source：arXiv Robotics — research abstracts · arxiv.org