Calibrated Ensemble-Based Chance-Constrained Safety for Autonomous Mobile Robots in Lean Intralogistics under Localisation Uncertainty

Autonomous mobile robots that run lean milk-run routes must keep a safe separation from operators, pallets and aisle boundaries while meeting a fixed delivery cycle. Their protective fields are usually sized once, for the worst localisation in the plant, so the vehicle stops at obstacles it could pass wherever it is well localised. We study uncertainty-aware safety filters for this setting, in particular an ensemble-based chance constraint that evaluates a discrete-time control barrier function on the members of an ensemble transform Kalman filter. Under a stated Gaussian assumption, the filter bounds the one-step probability that each separation requirement fails. We show that the constraint acts as a protective field whose margin is proportional to the localisation uncertainty in the direction of each barrier, and derive from it a local necessary condition on the pallet protrusion the vehicle can pass. Protective stops are applied only when a feasibility test shows that no certified command exists. In 4680 simulated runs of a four-station milk run with intruding pallets and crossing operators, a fixed margin sized for an occluded section gave a mean cycle time of 169.2 s, 66% of cycles on time and 17.2 operator calls per hour. The ensemble filter gave 127.8 s, 81% and 5.8 calls per hour, but 3.6 more separation events per hour, with 9 contacts in 10.7 simulated hours against 3 in 14.1 hours. At matched safety, all uncertainty-aware filters shared one frontier well below the fixed margin, and the ensemble filter showed no advantage over a linearised Kalman-filter chance constraint. The chance constraints differ mainly in calibration: their realised one-step risk followed the design level, whereas that of a covariance-scaled margin did not. The results are from simulation; the code, the data and a SymPy verification of the derivations accompany the paper.

Authors

Publication Details

Journal
Journal of Science Technology and Social Transformation
Published
2026-11-12
DOI
https://doi.org/10.64235/7f8v8883
Primary Topic
Robotic Path Planning Algorithms
Type
article
Field-Weighted Citation Impact
0.00
Controls
|||
ALL TIME
JAN
FEB
MAR
APR
MAY
JUN
JUL
AUG
SEP
OCT
NOV
article

Calibrated Ensemble-Based Chance-Constrained Safety for Autonomous Mobile Robots in Lean Intralogistics under Localisation Uncertainty

Toheeb Babatunde Ibrahim, Olasubomi Akanbi, Joseph Oladejo, Warith Akinade
Journal of Science Technology and Social Transformation
Robotic Path Planning Algorithms
article

Calibrated Ensemble-Based Chance-Constrained Safety for Autonomous Mobile Robots in Lean Intralogistics under Localisation Uncertainty

Toheeb Babatunde Ibrahim, Olasubomi Akanbi, Joseph Oladejo, Warith Akinade
article en

Abstract

Autonomous mobile robots that run lean milk-run routes must keep a safe separation from operators, pallets and aisle boundaries while meeting a fixed delivery cycle. Their protective fields are usually sized once, for the worst localisation in the plant, so the vehicle stops at obstacles it could pass wherever it is well localised. We study uncertainty-aware safety filters for this setting, in particular an ensemble-based chance constraint that evaluates a discrete-time control barrier function on the members of an ensemble transform Kalman filter. Under a stated Gaussian assumption, the filter bounds the one-step probability that each separation requirement fails. We show that the constraint acts as a protective field whose margin is proportional to the localisation uncertainty in the direction of each barrier, and derive from it a local necessary condition on the pallet protrusion the vehicle can pass. Protective stops are applied only when a feasibility test shows that no certified command exists. In 4680 simulated runs of a four-station milk run with intruding pallets and crossing operators, a fixed margin sized for an occluded section gave a mean cycle time of 169.2 s, 66% of cycles on time and 17.2 operator calls per hour. The ensemble filter gave 127.8 s, 81% and 5.8 calls per hour, but 3.6 more separation events per hour, with 9 contacts in 10.7 simulated hours against 3 in 14.1 hours. At matched safety, all uncertainty-aware filters shared one frontier well below the fixed margin, and the ensemble filter showed no advantage over a linearised Kalman-filter chance constraint. The chance constraints differ mainly in calibration: their realised one-step risk followed the design level, whereas that of a covariance-scaled margin did not. The results are from simulation; the code, the data and a SymPy verification of the derivations accompany the paper.

Journal of Science Technology and Social Transformation
Robotic Path Planning Algorithms
AI Navigator

Ask Laika to Summarize, Analyze, and Connect papers live on the map.

Summarize Papers & Methodologies

Extract key findings, datasets, and comparative methods across publications.

Benchmark Rankings & Visual Analytics

Rank top research institutions, authors, funders, topics, and journals by Field-Weighted Citation Impact (FWCI) and paper volume with instant charts.

Connect Distant Disciplines

Bridge topological clusters on the map to find hidden collaborative intersections.

Laika access coming soon