EON SPACE VENTURES · SIMULATION PROCEDURERobot approach recovery after a camera alignment fault
ESV-S04-SOP-02 · Robot servicing workcell · model v3 · 12 steps
Practise stopping and recovering a robot approach when camera evidence and satellite socket alignment disagree. Compare evidence, re-establish a reference pose and resume only after a reviewed correction.
Concept-model training and authoring specification. Real hardware operations require the programme owner’s approved procedure and engineering limits.
Before the exercise
Use authored robot motion and synthetic camera observations with clear provenance. Supply a reference view, approved approach envelope and stopped state. No autonomous perception accuracy or collision avoidance performance is assumed.
1 Load the approved approach state
Model labels: robot, satellite, sat.socket
Confirm the robot, satellite and socket identities. Load the saved reference pose and the procedure revision for the approach exercise.
Check: The correct asset and target must be established before motion.
2 Check the camera assembly
Model labels: robot.camera, robot.camera.lens, robot.camera.bracket
Identify the camera, lens and bracket. Compare their authored configuration with the reference view metadata.
Check: A camera label match alone does not establish calibration or image accuracy.
3 Inspect the viewing corridor
Model labels: robot.camera, gripper, sat.payload
Review the synthetic view corridor between camera, gripper and payload. Identify which object could occlude the target during the approach.
Check: The learner records a target visible condition from supplied evidence.
4 Move to the pre approach checkpoint
Model labels: robot.shoulder, robot.elbow, robot.wrist
Advance through the reviewed pose sequence to the pre approach checkpoint. Keep the gripper outside the socket envelope.
Check: No unverified inverse kinematics or physical joint limits are implied.
5 Compare the socket reference
Model labels: sat.socket, sat.socket.insulator, sat.socket.pins
Match the socket housing, insulator and pin area with the supplied reference image. Identify the declared alignment datum.
Check: The datum is author supplied; pin functions and electrical limits remain unknown.
6 Detect the conflicting observation
Model labels: robot.camera.lens, sat.socket, gripper
Present a reference overlay indicating alignment while the scenario pose record is offset. Stop and record the conflicting image and pose evidence.
Check: Continuing the approach on the overlay alone is a critical error.
7 Freeze and preserve the state
Model labels: robot.wrist, robot.forearm, gripper
Enter the authored stopped state and retain the current pose. Keep the gripper open and outside the restricted target envelope.
Check: The record distinguishes the stop event from successful recovery.
8 Return to the known checkpoint
Model labels: robot.shoulder, robot.elbow, robot.wrist, robot.camera
Follow the reviewed retreat path to the saved checkpoint. Compare the camera and wrist poses with their baselines.
Check: A blind direct move towards the target is not an accepted recovery.
9 Apply the reviewed correction
Model labels: robot.camera.bracket, robot.camera, sat.socket
Choose the supplied correction branch and update the synthetic reference alignment. Record the reviewer decision and the revised reference identifier.
Check: No learner invented calibration number or unreviewed offset is accepted.
10 Recheck visibility and alignment
Model labels: robot.camera.lens, gripper, sat.socket.insulator
Repeat the visibility check and compare the corrected reference with the pose record. Keep the target approach on hold until both agree.
Check: Unresolved or missing evidence keeps the robot at the checkpoint.
11 Resume the bounded approach
Model labels: robot.wrist, gripper.upper, gripper.lower, sat.socket
Advance to the authored final approach pose and stop before contact. Confirm the gripper and socket envelopes satisfy the training criterion.
Check: This exercise ends before insertion or a force controlled contact operation.
12 Release the recovery handover
Model labels: robot.camera, robot, satellite, sat.socket
Submit the original conflict, stop, retreat, correction and final approach evidence. Mark the superseded reference as unavailable for further steps.
Check: Completion requires a resolved discrepancy and a reproducible approved reference state.
Assessment and completion
Score reference identity 20, conflicting evidence recognition 25, stop and retreat sequence 25, corrected approach 15 and evidence provenance 15. Continuing through the fault or overriding an unresolved hold prevents completion.
Authoring and headless handoff
Use synthetic images and pose records labelled as training evidence. Do not claim live camera calibration, collision sensing, autonomous recovery or physical robot validation. Bind the actual GLB objects and retain reference revisions.
Reuse the exact GLB and part identifiers. Verify the supported Genesis import and binding path before generation. These are procedure specifications; they do not claim a completed Genesis run.