Organizations: State Key Laboratory of Transducer Technology, Aerospace Information Research Institute, Chinese Academy of Sciences, Beijing 100190, China · School of Electronic, Electrical and Communication Engineering, University of Chinese Academy of Sciences, Beijing 100049, China · School of Electronic and Optical Engineering,Zijin College,Nanjing University of Science and Technology,Nanjing 210023, China · School of Electrical and Automation Engineering, Nanjing Normal University, Nanjing 210023, China · School of Information and Communication Engineering,North University of China, Taiyuan 030051, Shanxi, China · School of Computing Science, University of Glasgow, Glasgow G12 8QQ, UK · Faculty of Electronic Engineering and Technology, Universiti Malaysia Perlis, Arau 02600, Malaysia
Mobile gas-sensing inspection requires a mobile platform to reliably reach ordered sampling poses. We present MIRA-PRM, a mission-informed probabilistic roadmap that combines geometry-, mission-flow-, and inspection-conditioned sampling with locally adaptive connectivity and evidence-gated refinement. Eight development campaigns comprising 142,448 planner runs evaluated mission scaling, repeated-query adaptation, finite node budgets, target distributions, local map repair, ablation, cross-family comparison, and parameter sensitivity. MIRA-PRM maintained 100% seed-level success as ordered targets increased from two to eight, and achieved 97.67% seed-level success under a 40-node ceiling, compared with 67.17% for PRM and 86.50% for PRM*. Local repair reduced event time by 79.9-81.9% and node additions by 90.8-94.3% relative to rebuilding. On a separate frozen holdout of 12 maps, 48 ordered missions, and three seeds per map-mission unit, MIRA-PRM produced 96/144 validated successes, versus 82/144 for PRM, 89/144 for PRM*, and 70/144 for LE-HG-PRM. On paired common-success runs, MIRA-PRM reduced wall time by 16.89% relative to PRM* and 81.01% relative to LE-HG-PRM, but was 52.23% slower than PRM; paths were 0.54-1.86% longer. All validated paths passed independent full-segment checks and clearance non-inferiority. The results support a reliability-oriented trade-off in structured simulations and motivate external-map and physical mobile-sensing validation.
Figures & tables
Figure 1: BESS-inspired structured industrial layouts used in the development experiments. The geometries represent regular rows, cross-zone traversal, a bottleneck, and temporary gate closures. Blue circles denote the deployment pose, coral diamonds denote ordered inspection poses, and dashed segments indicate the prescribed visit order rather than executable paths. These layouts support the Exp.1–8 development campaigns; the frozen holdout uses a separate set of procedurally generated maps.
Figure 2: Construction of the MIRA-PRM sampling distribution. Geometry support emphasizes constrained free-space bands; mission-flow support follows conservative guides between ordered sampling poses; inspection support concentrates around required poses; and the final density combines the active normalized fields with a 15% uniform exploration component. Intensity indicates relative sampling density, not the probability that a path follows a particular corridor.
Component
Information used
Functional role
Geometry prior
Constrained free-space bands
Improves representation of gates and narrow passages.
Mission-flow prior
Ordered guide paths across all mission legs
Allocates persistent capacity to repeatedly demanded corridors.
Inspection support
Clearance-adaptive support around required poses
Supports endpoint connectivity without duplicating strong mission-flow mass.
Adaptive connectivity
Larger local k in constrained regions
Compensates locally for collision-rejected neighbours.
Evidence-gated refinement
Disconnection, support, detour and deviation diagnostics
Adds nodes only where route-level evidence indicates insufficient graph support.
Table 1: MIRA-PRM mechanisms and their functional roles.
Design item
Value
Topology families
Maze; rooms; random obstacles
Core maps / missions
12 maps / 48 ordered missions
Inspection targets
6 per mission
Planner seeds
3 per map–mission unit
Core algorithms
MIRA-PRM; PRM; PRM*; LE-HG-PRM
Core resident ceiling
120 nodes
Table 2: Frozen evaluation design.
Map family
MIRA-PRM
PRM
PRM*
LE-HG-PRM
Maze
2/48 (4.17%)
0/48
0/48
0/48
Rooms
46/48 (95.83%)
34/48 (70.83%)
41/48 (85.42%)
22/48 (45.83%)
Random obstacles
48/48 (100%)
48/48 (100%)
48/48 (100%)
48/48 (100%)
Table 3: Validated seed-level success by topology; each denominator is 48.
Endpoint
Baseline
MIRA effect [95% CI]
Holm adjusted p
Interpretation
Seed-averaged success
PRM
+9.72 pp [3.47, 15.28]
0.2500
Secondary positive estimate
Seed-averaged success
PRM*
+4.86 pp [1.39, 9.03]
0.2500
Secondary positive estimate
Seed-averaged success
LE-HG-PRM
+18.06 pp [12.50, 26.39]
0.1875
Secondary positive estimate
Observed robust 3/3
PRM
+14.58 pp [4.17, 22.92]
0.5000
Secondary; not 5/5
Observed robust 3/3
PRM*
+6.25 pp [0.00, 12.50]
0.5000
Secondary; not 5/5
Observed robust 3/3
LE-HG-PRM
+22.92 pp [16.67, 29.17]
0.3750
Secondary; not 5/5
Table 4: Pairwise effects from the frozen holdout. Positive percentages favour MIRA-PRM.
Figure 3: Paired outcomes in the frozen holdout (Exp.9). Dots show comparator and MIRA-PRM means over the same map–mission units. Success rate and PAR-2 retain all 48 units per comparison; wall time first uses exactly seed-paired common-success outcomes, takes the within-mission median, and then averages the available missions. Thus, the MIRA-PRM wall-time mean can differ by comparator. n denotes paired missions. These are descriptive means, not the relative-effect estimates in table 4 . Cyan dots denote MIRA-PRM and pale-green dots denote the comparator.
Figure 4: Distributions of paired mission-level effects in Exp.9. Each point is one map–mission comparison; half-violins show density and boxes show the median and interquartile range. Positive values indicate lower MIRA-PRM time or path length. Time and path effects use exactly seed-paired common-success outcomes aggregated within mission; PAR-2 includes failures. n denotes finite paired missions. The observations are descriptive; inferential intervals and tests account for map clustering as reported in table 4 .
Algorithm
Success (%)
Robust success (%)
Time (s)
A*-reference ratio
MIRA-PRM
100.00
100.00
0.06533
1.01527
PRM
99.86
98.89
0.03826
1.04989
PRM*
100.00
100.00
0.08999
1.02591
LE-HG-PRM
100.00
100.00
0.45061
1.02722
HG-PRM
99.72
97.78
0.29175
1.03670
RRT
100.00
100.00
0.03776
1.11373
Table S1: Cross-family comparison under the configured settings in Exp.7.