Working project / Robotics vision & sensing
A moving target.
A visible assumption.
How do occlusion and delayed measurements shift a conveyor pick target? Explore the question with a deterministic, synthetic model—not a real camera or robot.
Conveyor Sensing Bench / Occlusion & Latency
Synthetic · AI-assisted educational prototypeA rectangular part moves at a known constant speed. A delayed camera-like measurement sees only its left portion. Compare that perturbed visible-centroid target with a correction that is given the exact geometry, occlusion and speed.
Change one assumption.
Starts paused. Play stops at 3 seconds. Changing a parameter pauses playback. Sliders support arrow keys, Home and End.
Ready. No animation is running.
Inspect the model, current values & limitations
Model: x(s) = 140 + v·s mm, defined for −0.4 ≤ s ≤ 3.2 seconds. The displayed time t runs from 0 to 3 seconds. Width W = 80 mm; arrival horizon H = 0.20 s.
Visible interval at capture time t − L: [x(t − L) − W/2, x(t − L) + W/2 − fW]. Its centroid is x(t − L) − fW/2.
Disturbance: e(s) = A·sin(2π·0.7s + 0.3) mm. This deterministic bounded signal is not a Gaussian noise model, standard deviation or confidence interval.
Raw target: z = x(t − L) − fW/2 + e(t − L).
Future truth: g = x(t + H). Corrected target: c = z + fW/2 + v(L + H). Therefore c − g = e(t − L), up to floating-point rounding.
The correction removes the imposed geometry/timing bias under exact assumptions. It does not always reduce the instantaneous absolute error: the disturbance can partially cancel raw bias.
Limits: exact width, occlusion fraction, timestamps and constant speed are supplied—not estimated. No pixels, calibration, acceleration, rotation, depth, grasp mechanics, collision checking, trained vision, real sensor data or robot commands are involved. There is no accept/reject safety threshold and no industrial performance claim.
Why this question matters: it makes the distinction between a visible feature and a time-aligned target inspectable. A real system would need measured inputs, independently tested estimation and hardware-specific validation.
Synthetic educational model · AI-assisted implementation · No camera, hardware or robot control · Not safety-certified
Interactive values require JavaScript. Equations and limitations remain available above.