Centria University of Applied Sciences

Robots executing in unconstrained environments frequently must select actions on an incomplete knowledge basis and yet satisfy geometric and physical feasibility constraints. Hybrid conditional planning manages this tension by constructing a conditional plan tree whose branches are guarded by feasibility checks. Full expansion of such trees for offline execution can be prohibitively expensive, and assumes no deviation from the generated branches will occur at execution time. Fully online replanning responds to failures without pre-computation, but often re-plans repeatedly from scratch without exploiting previouslycomputed contingencies. We present Reactive-HCPlan, an online-offline hybrid conditional planner which considers the conditional plan as a lasting structure that is kept current during execution. An initial conditional structure is generated offline with a reactive depth threshold criterion, and observed deviations during execution are monitored for the divergence of observed from modeled outcomes and from physical/geometric feasibility limits. When a discrepancy is found, ReactiveHCPLAN selectively repairs the structure in only the failing portion (e.g., adds an unmodeled observation branch or repairs a failing subtree). We simulate the approach on the kitchen table object manipulation benchmark with disturbances including object displacements, occlusion-induced changes to sensor inputs, grasping failures, and motion plan blockages. Across these scenarios, Reactive-HCPlan yields markedly smaller conditional structures than complete offline planning and reduces repeated re-planning relative to an execution monitoring baseline, largely due to reuse of validated plan fragments.

Structured electronic patient care records can support consistent documentation, information transfer and secondary data use, but implementation depends on users’ competence and system compatibility with clinical work. Evidence concerning education and competence development for structured documentation in prehospital emergency care remains limited. This secondary analysis examined paramedics’ education and development needs and whether assessments of KEJO training adequacy, timing and documentation competence differed by age, educational background or the electronic patient care record previously used. Data originated from an anonymous cross-sectional web survey conducted in one Finnish wellbeing services county in February–March 2024. Responses from 137 professionals (response rate 30.4%) were examined using descriptive statistics and non-parametric group comparisons, contextualised by previously reported findings and an expanded inductive analysis of free-text responses. Overall, 46.7% considered theoretical instruction sufficient, and the same proportion considered face-to-face training sufficient. Most respondents (83.9%) considered the training-to-implementation interval too long, while 54.7% assessed their documentation competence positively. After Holm adjustment for multiple comparisons, no statistically significant differences were detected across the background groups. Qualitative findings emphasised hands-on practice, authentic cases, simulation, refresher training, concise update materials and timely support. Respondents distinguished educational needs from problems requiring system development. The comparisons provided no evidence of systematic differences between the groups examined. Training should be coordinated with implementation and user-centred system development as a continuous process rather than a single pre-launch event.

Time-Aligned Peaks (TAP) is a Python tool for analysing and visualising peak events across multiple time series. TAP pairs a conventional line plot with a stacked peak-presence timeline sharing the same x-axis, making co-occurring peaks immediately visible while preserving alignment. The pipeline ingests tabular data, performs unified timeline alignment with configurable resampling and missing-value policies, detects peaks using a simple slope-change criterion and exports reproducible artefacts: a combined image, a timestamped peak report and a binary peak matrix. TAP supports fast, reliable inspection of multi-series peak dynamics in small-to-medium datasets with single-command reproducibility.

The increasing complexity of computational datasets demands practical solutions for visualizing n-dimensional arrays in a consistent and accessible format. This work presents an automated and unified procedure that converts bitmaps and other structured rectilinear array data with any non-negative, integral number of spatial dimensions into a unified two-dimensional (2D) grid. It expands 0D and 1D data into 2D planes while reducing 3D and higher-dimensional data into 2D projections, enabling simultaneous visualization of all planes (slices) for intuitive navigation. Key features include global colormap scaling for consistency across slices, automated slice annotation with indexing, alternating background colors for clarity in higher dimensions, and optional autoscaling for readable text. Benchmarking against established tools demonstrates improved visualization completeness, readability, and navigability, particularly for large arrays. The lightweight, open-source Python implementation integrates directly into workflows without custom coding or heavy dependencies. Accessibility considerations are prioritized through perceptually uniform colormaps that balance clarity and inclusivity. The method has been successfully applied to solve and verify solutions to challenging computational geometry problems and is broadly applicable across various fields, including robotics, biology, manufacturing, and architecture. By addressing limitations of existing visualization tools, the proposed approach directly supports computational science by bridging algorithmic solutions and human-readable visualization, integrating into modeling workflows, and providing an open-source implementation for reproducible analysis. Supplementary materials include full implementation, usage examples, benchmarking datasets, and validation results to facilitate immediate adoption in research and applied settings.

The electrification of road transportation poses multi-faceted challenges in the Nordic region, where cold ambient temperatures, long distances, and limited charging infrastructure affect operational efficiency and feasibility. This study presents a novel integrated framework that combines artificial intelligence, digital twins, and co-simulation to support energy-efficient routing and infrastructure planning for heavy-duty electric vehicles in the Nordic region. A Gradient Boosting model was trained on over 180 real-world intercity trips of data collected from an electric truck to estimate battery depth of discharge based on key operational factors. Motor load, torque, and total weight emerged as dominant predictors, while ambient temperature had a moderate effect. A digital twin environment was developed to assess the impact of various charging infrastructure scenarios, and a real-time co-simulation model was used to evaluate energy flow and support cost analysis. Results confirm the framework’s capability to support infrastructure planning and fleet operation optimization during winter. The approach provides actionable insights for logistics operators and policymakers who are advancing the electrification of heavy-duty vehicles in the Nordic region.