TIME TO NEXT EXPERIMENT
RFR
Failure Reconstruction Engine
Spend less time debugging. Spend more time validating value.
RFR connects device communication inside a robot, embedded Linux, LTE, and cloud evidence on a shared timeline to shorten the time it takes physical AI teams to return to their next valuable real-world experiment.
Visit the RFR public website ↗RFR
TIME TO NEXT EXPERIMENT
Spend less time debugging. Spend more time validating value.
CHALLENGE
The problem
Robot development teams often lose time to protocol discovery, log preparation, waiting for failures to recur, clock alignment, and handoffs between teams. When logs from the robot to the cloud use different formats and clocks, narrowing the scope of a failure becomes especially difficult and real-world value validation stops.
APPROACH
Current approach
RFR normalizes evidence from devices, ROS 2, Linux, modems, LTE, transport, cloud systems, and recovery processes into common events, aligns their clocks, and reconstructs a Failure Timeline. It separates the first observed deviation, evidence, uncertainty, and missing evidence to guide the next capture or check.
SYSTEM DIAGRAM
FAILURE RECONSTRUCTION
Reconstruct a failure across the full system
Evidence from different formats and clocks is normalized before the failure scope is narrowed.
DEVICE
Sensors / controllers
ROBOT
ROS 2 / embedded Linux
NETWORK
Modem / LTE / transport
CLOUD
API / service / recovery
NORMALIZE
Common event model
Preserve raw evidence and align clocks
FAILURE TIMELINE
Deviation → evidence → uncertainty
Separate facts from missing evidence
CONTINUE
Next capture or experiment
Return the team to value validation
RESEARCH FLOW
Research and validation flow
- 01
Connect
Understand the communication path and connect the equipment and required observation points.
- 02
Detect
Detect anomalies and preserve evidence from before and after a failure.
- 03
Reconstruct
Align different formats and clocks, then narrow the failure scope using evidence.
- 04
Continue
Use the findings to return to the next valuable real-world experiment.
CURRENT STAGE
Current research stage
The current v0 starts with one resolved mobile connectivity failure reconstructed from three to five available log sources. Connectivity Failure Reconstruction Sprints and connectivity or wireless PoCs for existing equipment are being validated jointly with partners.
About this information
This page summarizes information published on the RFR public website as an invilab proprietary R&D project. Public information reviewed: August 10, 2026.
The project is under research and development. Publication does not guarantee future commercial availability or implementation of every described capability.