Search Results (273)

Personalized learning pathways are difficult to support in distributed educational systems because learner records cannot always be centralized and one shared federated predictor may not represent heterogeneity across institutions, learning tasks, and individual learners. We propose FedPath-MTL, a federated multi-task framework with global, task-specific, and learner-specific parameters, a bounded future-error proxy, a dynamic task graph, reliability-aware aggregation, and a constrained primal–dual beam-search planner. We evaluate only the next-outcome predictive component on four public educational releases: ASSISTments 2009–2010 Combined, Junyi Academy 2018–2019, EdNet-KT1, and OULAD. Every method is evaluated both before and after the same validation-only temperature-scaling protocol, and AUC, negative log-likelihood (NLL), Brier score, and 15-bin expected calibration error (ECE) are reported. The pooled central comparator has the highest mean AUC on all four releases. FedPath-MTL AUC ranges from 0.531 to 0.580, and identically calibrated ECE ranges from 0.098 to 0.127; matched comparisons with FedAvg-style training do not show a uniform AUC or calibration advantage. Observed school identifiers define clients only for ASSISTments; the other releases use explicitly labeled simulation partitions. The public logs do not jointly identify actions, propensities, subsequent-learning rewards, resource catalogs, and institutional constraints, so no pathway-effectiveness or causal learning-gain claim is made. Software tests verify equation-level implementation of the planner but do not validate educational benefit or real-institution constraint satisfaction. Full article

UAV pursuit–evasion in multi-obstacle environments constitutes a safety-critical decision-making problem in which the pursuer must intercept a maneuvering evader while satisfying obstacle-avoidance constraints. Existing safety-constrained methods typically optimize expected safety costs or apply immediate interventions, but they do not explicitly determine when recovery should begin once finite-horizon risk emerges. To address this issue, we propose the ReSwitch framework, which separates task-oriented pursuit from safety-oriented recovery. When flight risk emerges, a value-preserving switch-time planner evaluates candidate switching times through opponent-conditioned hybrid rollouts. The feasible switch time with the largest pursuit value after recovery is selected for execution, allowing the controller to prioritize safety while preserving pursuit effectiveness whenever possible. Experiments demonstrate that ReSwitch achieves a mean success rate of 89.02% and a physical safety rate of 91.67%, showing favorable performance compared with the baselines. These results indicate that ReSwitch provides a favorable pursuit–safety trade-off under the tested conditions. Full article

Reliable UAV navigation is challenging when direct geometric perception is limited and only sparse environmental observations are available. This paper presents CoSafe-Nav, an integrated navigation framework that uses ultra-wideband (UWB) line-of-sight (LOS) observations to construct a candidate corridor and guide trajectory generation. The framework addresses fixed-altitude planar navigation with known, pre-deployed anchors and an available UAV pose estimate. LOS-supported segments are accumulated online, and a grid-based Euclidean distance transform (EDT) provides a common clearance representation for path ranking and trajectory refinement. The path planner combines travel distance with a local inverse-distance penalty. The trajectory stage corrects interior control points using the EDT gradient, regenerates the spatial curve, and assigns execution timing. Evaluation comprises two simulation scenarios, component comparisons, an anchor-availability study, and indoor UAV trials. CoSafe-Nav completed all ten navigation tasks in each simulation scenario and four of five indoor trials. In the path-planning comparison, mean EDT corridor clearance increased from 0.76 to 0.98 m and from 0.63 to 0.91 m, accompanied by longer routes. These means describe successful tasks within each configuration. The trajectory comparison also showed a lower acceleration integral for the safety-guided method. The results support the feasibility of the integrated processing chain under the stated pre-instrumented deployment conditions. Full article

Autonomous navigation across heterogeneous environments remains challenging because different environments may require fundamentally different planning representations and action models. Structured roads are naturally represented as graphs with constrained connectivity, unstructured terrain requires continuous geometric reasoning, and service-specific areas such as charging stations can be described by discrete task states and symbolic actions. We propose a modular, context-aware planning framework that represents the environment as a set of navigation contexts, each with its own environment representation, planning state space, applicable actions, and dynamics. Contexts are connected through explicitly defined interfaces that determine where inter-context transitions can occur and how states are mapped between the corresponding representations. Planning is formulated as an incremental A*-style search over a joint state space comprising the current context, context-specific state, and internal resource variables such as battery level. The planner jointly considers intra-context actions and inter-context transitions while minimizing accumulated action and context-transition costs subject to modeled resource constraints. We evaluate the framework in simulated environments combining continuous geometric, graph-based, and discrete task-level representations. Compared with a hierarchical multi-context planning baseline using fixed representative interface states, the proposed method produced shorter routes, retained more battery at the goal, and required less planning time in both evaluated resource scenarios. In a restricted grid–topological evaluation, both the proposed method and a strengthened topological baseline found the shortest path in all 50 test cases, while the proposed method achieved lower average planning time. A nearest-node topological baseline frequently failed or produced suboptimal paths. Full article

This paper addresses low-altitude navigation of multiple UAVs through a shared two-dimensional environment with narrow passages, blocked cells, and predicted moving obstacles while preserving an assigned task order. The planner receives ordered task bundles from an allocator and checks static obstacles, moving-obstacle timing, sampled minimum inter-UAV separation, deadlines, risk budgets, and an energy proxy. It generates risk-weighted candidate path segments, repairs timing conflicts with waits and local detours, verifies service and terminal occupancy, and returns a certificate that records whether a segment is executable, its total travel cost, risk exposure, and energy proxy, or the reason for failure. We compare no feedback, context no-good, typed-failure, quantitative, and combined feedback under medium-load and high-stress test suites. Quantitative feedback lowers risk per completed task in both suites after correction for multiple comparisons. Failure-type feedback adds no detectable benefit, and completion-rate differences do not remain significant after the same correction. Fixed-bundle simulations show that the proposed planner can preserve scheduled executability while reducing threat exposure relative to a spatiotemporal-priority baseline. Single-UAV flights demonstrate waypoint execution, reference tracking, and avoidance of designated regions. Full article

The Newton Downhill Optimizer (NDO) combines a derivative-free downhill relation with population differences. However, its Hybrid-Guided Operator uses a random reference and persistent best-solution guidance, which can cause directional fluctuations and premature population contraction. This study proposes the Adaptive Fitness–Distance-Guided Newton Downhill Optimizer (AFDNDO). Fitness–Distance Balance selection identifies guiding individuals that account for both solution quality and spatial diversity. Stage protection, elite protection, and historical success-rate feedback regulate activation of the improved branches. A tripodal heavy-tailed update and a wave-weighted masked differential update reconstruct the two original branches. A non-uniform mutation is also triggered for low-quality individuals when the global best value stagnates. Across 30 independent runs on CEC2017, CEC2020, and CEC2022, AFDNDO attained the lowest mean rank in all six formal benchmark configurations. Its mean ranks on the 10-, 30-, and 50-dimensional CEC2017 tests were 1.172, 1.241, and 1.276, respectively. Dynamic path-planning environments included rigid obstacles, three levels of soft-risk regions, and moving obstacles. In the single-target environments, AFDNDO–DWA achieved a 100% execution success rate without collisions. In the multi-target environments, it reduced the mean objective value by 4.46–6.64% relative to NDO. It also increased the success rate from 63.33% to 70.00% in the most constrained environment. These findings indicate that AFDNDO improves cross-landscape optimization performance while retaining the basic NDO framework. They also support its use as a global planner within the tested dynamic multi-task environments. Full article

Background: Large language models (LLMs) can produce clinically plausible recommendations that are not adequately supported by authoritative evidence. Objectives: We introduce GuidelineGuard, a modular multi-agent retrieval-augmented generation pipeline in which a separate Auditor verifies claim–sentence support before a recommendation is surfaced. Methods: The original evaluation used a 73-sentence guideline corpus and GG-Bench-60, with replication across three open-weight backbones. In response to reviewer concerns about benchmark size and selective evaluation, we added a source-traceable GG-Bench-200 stress test and the complete 500-case held-out PQA-L test split of PubMedQA. The revision experiments compare single-pass RAG, a paired multi-agent no-Auditor ablation, and GuidelineGuard; the paired runner is designed to share the Planner–Retriever–Clinician draft so that the Auditor is the only intervention. Checkpoint verification confirmed an identical observable pre-audit state for all 200 GG-Bench cases and 496/500 PubMedQA cases; four PubMedQA cases were regenerated after quota-interrupted resumption and were correct commitments in both arms. Because the originally used hosted Llama endpoints became unavailable after the initial experiments, the expanded runs use openai/gpt-oss-20b for generation and openai/gpt-oss-120b for the Auditor. Results: On GG-Bench-200, single-pass RAG achieved 0.970 operational accuracy, while the no-Auditor and GuidelineGuard arms achieved 0.955 and 0.925, respectively. GuidelineGuard committed on 186/200 cases (coverage 0.930) and was correct on 185/186 commitments (selective accuracy 0.995); all 186 commitments cited at least one gold evidence identifier. Relative to the paired no-Auditor arm, the gate rejected six otherwise-correct commitments and no incorrect commitment. On PubMedQA-500, single-pass RAG achieved 0.644 operational accuracy at 0.950 coverage, the no-Auditor arm 0.638 at 0.896 coverage, and GuidelineGuard 0.550 at 0.736 coverage. Selective accuracy increased across those operating points from 0.678 to 0.712 to 0.747. Within the 496 PubMedQA cases with verified-identical observable pre-audit state, the gate rejected 36 incorrect and 45 correct pre-audit commitments, demonstrating both error enrichment and a substantial false-rejection cost. Conclusions: The expanded results support GuidelineGuard as a selective claim–evidence verification mechanism, not as a universally more accurate generator. Its value is the explicit, auditable coverage–risk trade-off; the appropriate verification threshold is task- and cost-dependent and requires prospective clinical validation. Full article

Large-area remote-sensing data provide essential pre-mission information for unmanned ground vehicles, but their spatial support and temporal latency may obscure local terrain changes. A remaining challenge is to translate heterogeneous regional evidence and recent local observations into a consistent, updateable, and planner-ready map. This study presents a satellite–unmanned aerial vehicle (UAV) workflow for constructing and incrementally maintaining an off-road traversability map for mission-level global planning. A common H3 index organizes satellite imagery, terrain, soil, road evidence, and local UAV semantic observations while retaining their native spatial support and provenance. The map separates environmental-prior, semantic, and traversability-cost layers to support interpretable fusion and independent updating. A confidence-hierarchical conflict resolution mechanism resolves inconsistencies in the regional prior, while an observer-agnostic interface projects UAV semantic observations onto local map cells. RGB imagery is used by the primary UAV observer, and digital surface model (DSM) is evaluated as an optional semantic-observation modality. Evaluation included a manually reviewed regional benchmark, a unified buffered spatial holdout, cell-level update assessment, and 40 fixed replanning tasks. Conflict resolution reduced high-risk omissions. RGB-only SegFormer-B2 achieved the highest semantic accuracy with moderate computational complexity. UAV override achieved a cell-level F1 score of 96.96% and limited the false-positive accumulation associated with conservative union. Replanning further revealed a trade-off between hazardous-cell avoidance and search-graph connectivity. The proposed workflow provides a maintainable interface between multi-source remote sensing and global UGV planning rather than a replacement for onboard perception, local obstacle avoidance, or vehicle control. Full article

The lunar south polar region’s extreme illumination conditions impose strict energy constraints for solar-powered rover operations. Traditional Sun-synchronous path planning relies on dynamic time-dependent illumination evaluation, leading to high computational costs. We present CIRsE-Net, a spatiotemporal deep learning model that generates 72 h continuous illumination maps from hourly sequential illumination data. The model integrates a lightweight SST-VGG encoder (a customized 14-layer CNN for spatial feature extraction), BiGRU temporal modelling (a bidirectional recurrent network for capturing forward and backward temporal dependencies), and a consistency-aware spatiotemporal attention mechanism. On three different lunar illumination datasets (20 m/pixel, 5 m/pixel, and 20 m/pixel with a 2 m panel height), the model achieves Dice scores up to 0.983 and accuracy up to 0.985. When integrated with an enhanced 3ST-A* planner, the framework converts dynamic path planning into a static search task, reducing computational overhead while preserving path optimality and satisfying slope and illumination constraints. This work provides a validated methodological framework for Chang’E-7 mission planning and future lunar polar exploration missions by transforming dynamic path planning into a static search task. Full article

Census-style address records contain heterogeneous structures that must be decomposed into fine-grained fields before they can support linkage, geocoding, and administrative processing. This paper describes the design and implementation of a three-stage Planner–Manager–Worker pipeline using a locally deployed open-weight language model. The prototype includes address-specific extraction, same-model semantic consistency review, formatting, record-preserving chunking, and an exploratory bounded retry pathway. Although the prototype also implements workers for names, phone numbers, email addresses, Social Security numbers, and dates of birth, the quantitative evaluation is limited to address-component parsing. We conduct a controlled pilot using 700 author-generated synthetic records distributed across seven address categories. Each record was first created as a structured object and then rendered into text; the original structured object served as the evaluation reference. Across three repeated executions on the same fixed benchmark, the complete hierarchical configuration obtained a mean micro-averaged address-component exact-match score of 95.7%, compared with 80.9% for the evaluated monolithic LLM reference; the deterministic rule-based reference obtained 52.3%. The fixed prompts were not matched, and realized inference calls, dynamic token allocation, original batching, and reported-run segmentation could not be verified from retained artifacts. No component ablations were conducted. Consequently, the numerical difference cannot be attributed uniquely to planning, task coordination, specialization, semantic review, feedback, chunking, or multi-agent organization. The reference labels were not independently annotated, the benchmark does not estimate performance on naturally occurring records, and the retry pathway was invoked only three times. The contribution is therefore the specification and controlled pilot characterization of an on-premise address-parsing pipeline, rather than evidence of architectural causality, production readiness, broad PII-extraction accuracy, or real-record robustness. Full article

Canopy occlusion, dynamic vegetation, structural degeneracy, and implicit terrain risks make stable localization and task-adaptive planning difficult for resource-constrained orchard robots. This paper proposes CitraNav, a lightweight navigation method for global navigation satellite system (GNSS)-denied orchards. It separates stable geometric evidence for localization from short-lived semantic evidence for planning, preventing semantic observations from accumulating in the global map. For localization, a hierarchical voxel map selects its resolution according to local structure and light detection and ranging (LiDAR) sampling characteristics, while cross-frame reliability and observability constraints suppress updates from transient vegetation and weakly observable directions. For planning, synchronized color and depth observations form a local semantic risk point cloud. A model predictive path integral (MPPI) planner combines task-dependent semantic costs with exact-footprint collision checking against currently detected obstacles. In simulation, CitraNav achieved a mean translational localization root mean square error (RMSE) of 0.075 m. Compared with geometric point-cloud planning, semantic planning reduced the collision rate by 71.4% and increased weed coverage 4.72-fold. Across 14 real-world sequences spanning farm-road, lawn, forest, and orchard environments, CitraNav achieved mean translational and heading RMSEs of 0.151 m and 1.13°, respectively, while using 72.3–87.4% fewer geometric map cells than the comparison methods. The complete perception–planning pipeline operated at 20.3–32.7 frames per second on an edge platform. These results suggest that CitraNav offers a balanced approach to localization stability, task-adaptive planning, and computational efficiency in complex orchard navigation. Full article

Geospatial data provide an important basis for urban governance, resource management, disaster assessment, and public health analysis by linking spatial locations, attribute information, and dynamic processes. However, complex geospatial analysis still requires substantial expertise in spatial databases, spatial SQL, and GIS computation tools. Although large language model-based Text-to-SQL methods have lowered the barrier to natural language-driven data querying, most existing approaches rely on single-step SQL generation and remain unstable for spatial tasks that involve attribute retrieval, spatial relationship evaluation, geometric operations, and statistical aggregation. To address this limitation, this paper proposes a Code Generation-Driven Query–Computation Separation Framework (CGD-QCSF). The framework is based on the separation of query and computation, and decomposes complex geospatial analysis into a staged execution process. CGD-QCSF coordinates intent understanding, schema pre-filtering, planning, execution state management, SQL generation, and spatiotemporal computation. A structured planner and an execution state manager coordinate task decomposition, capability-aware routing, and evidence-based recovery. A SQL Code Generation Agent (SCGA) handles database access, attribute filtering, and intermediate data extraction, while a Spatiotemporal Computation Agent (STCA) performs out-of-database spatial computation and statistical aggregation in an isolated Python sandbox. We construct a benchmark of 200 tasks, covering easy, medium, and hard spatial tasks. In the main experiment with Qwen3.7-Plus as the foundation model, CGD-QCSF achieves a Strict Structured Accuracy (SSA) of 90.5%. Removing the Planner reduces SSA to 84.5%, while removing the STCA reduces it to 70.5%. The ablation experiments show that removing either the Python sandbox or the Planner Agent degrades performance on complex tasks. These results indicate that CGD-QCSF extends complex geospatial analysis from single-step SQL generation into a multi-staged execution process. By explicitly separating query and computation, the framework reduces interference between spatial computation logic and database schema information, thereby improving the stability and success rate of natural language-driven geospatial analysis. Full article

In heterogeneous multi-robot systems, allocating tasks efficiently requires determining whether each robot can actually carry out a given task. In existing multi-robot task allocation research, however, such task feasibility has typically been handled inside a particular optimizer or symbolic planner, while spatial traversability has been assessed against static criteria that cannot capture the changes induced by a robot’s loaded state. This paper proposes an ontology-based semantic feasibility reasoning method for heterogeneous multi-robot task allocation. The proposed method defines semantic models for robots, tasks, and places and applies hybrid reasoning, combining declarative reasoning with procedural evaluation, to determine multi-axis capability conditions and loaded-state place reachability. The reasoning result is formalized as an allocator-independent ReasonerOutput that serves as a common input for diverse allocation algorithms. In experiments spanning four scenarios over three fleet configurations and four allocators, together with an ablation study on 200 randomized instances at each of three problem scales, the proposed ReasonerOutput consistently functioned as a shared semantic feasibility constraint. The experiments further showed that both the fleet composition and the loaded state of the target item affect assignment feasibility. These results indicate that, for the static one-shot assignment setting evaluated here, the proposed method makes the task feasibility of heterogeneous robots explicit through semantic reasoning and allows allocators of differing algorithmic character to draw on this feasibility in common. Full article

Permanent lunar habitation will require robotic systems that can maintain infrastructure, recover from local failures, and acquire new operational capabilities under limited Earth supervision. Existing planetary robots are largely fixed-function specialists, while end-to-end foundation-model policies remain difficult to validate and extend for safety-critical surface operations. We present SELENE (Self-Evolving Lunar Embodied ageNt Ecosystem), an architectural proposal for contract-governed lunar robotic autonomy centered on a shared Atomic Action Library $\mathbb{A}$. The key abstraction is the Atomic Action Contract: a typed skill interface that specifies parameters, preconditions, goal predicates, execution bindings, safety envelopes, runtime reports, and validation metadata. Through this contract, a VLM-driven Cognitive Agent plans over executable skills, a multi-modal Execution Agent realizes them through optimization-based controllers, Vision–Language–Action (VLA) policies, Vision–Language–Navigation (VLN) policies, or reinforcement-learned policies, and an offline Evolutionary Agentic Framework synthesizes and registers new candidate contracts without modifying the planner or the execution interface. This paper presents an architecture-level validation of that contract mechanism. We instantiate SELENE across two heterogeneous pathways on LunarBot and its simulation counterpart, with optimization-based control supported as a third execution modality. A pre-trained VLA policy adapted from 100 teleoperated demonstrations achieves 29/30 task success (96.7 percent) in in-domain trials on the physical LunarBot. A curriculum–RL policy instantiates the traversal pathway in simulated lunar-gravity terrain. Together, these results show that the Atomic Action Contract can serve as a common registration and dispatch interface across heterogeneous control modalities. The same contract layer also defines the path toward runtime gap-triggered self-evolution, mission-grade admission, and lunar-environment validation in subsequent system-level studies. Full article

Autonomous micromobility vehicles (AMVs) need local planning that balances task progress, safety, and pedestrian interaction quality in pedestrian-rich shared spaces. Evaluating such planners is difficult because studies vary scenarios, seeds, metrics, outputs, and aggregation rules, while single-score leaderboards hide which behaviors produce a ranking. This paper proposes a repeatable, auditable, multi-objective benchmark protocol for AMV local planning. Before comparison, it fixes the scenario set, repeated seeds, measured metrics, stored outputs, planner-interface records, and aggregation procedure. We demonstrate it with a frozen robot_sf_ll7 campaign: 47 shared-space scenarios, three evaluation seeds, and 141 scenario-seed episodes per planner in one differential-drive AMV configuration. The stress test surfaces a descriptive safety–performance separation: a Proximal Policy Optimization (PPO)-family profile reaches higher observed mean task success than a classical reciprocal-avoidance baseline, while that baseline keeps lower collision exposure. Absolute success stays low for both, with most scenarios unsolved by either. Because this learned policy was trained on a superset of the evaluation scenarios, its higher success reflects behavior on the benchmark set, not held-out generalization—an overlap the protocol records per planner rather than hiding in one score. The finding is bounded to these configured pipelines, not a universal planner-family ranking. The contribution is an auditable comparison framework tracing results from the scenario matrix and fixed seeds to episode records, aggregate reports, and manifests. Full article