Mengke Zhu | Searcharxiv

arXiv2026

Bridging Perception and Action: A Lightweight Multimodal Meta-Planner Framework for Robust Earth Observation Agents

Autonomous Earth Observation (EO) agents are transitioning from passive perception to complex, multi-step task execution. However, current architectures that integrate planning and execution within a single model often struggle with combinatorial complexity and reasoning errors in dynamic EO scenarios. To resolve these challenges, we propose the Lightweight Multimodal Meta-Planner (LMMP) framework. LMMP incorporates a dual-awareness mechanism that grounds strategic plans in both multimodal image features and high-level task semantics. Crucially, we introduce a Meta Task Library to inject remote sensing expert knowledge directly into the workflow, which standardizes domain logic and ensures plans are physically feasible. We further implement a two-stage training pipeline, initializing the Meta-Planner via expert-distilled Supervised Fine-Tuning and refining it through Direct Preference Optimization based on execution feedback. Extensive experiments on a dataset derived from EarthBench and ThinkGeo demonstrate that LMMP significantly improves tool-calling accuracy and task success rates. Moreover, the framework exhibits strong ``plug-and-play'' versatility, consistently enhancing the performance of diverse executor backbones across previously unseen EO missions.

cs.MA↗

arXiv2025

Autonomous Guidewire Navigation for Robot-assisted Endovascular Interventions: A Knowledge-Driven Visual Guidance Approach

Autonomous robots for endovascular interventions hold significant potential to enhance procedural safety and reliability by navigating guidewires with precision, minimizing human error, and reducing surgical time. However, existing methods of guidewire navigation rely on manual demonstration data and have a suboptimal success rate. In this work, we propose a knowledge-driven visual guidance (KVG) method that leverages available visual information from interventional imaging to facilitate guidewire navigation. Our approach integrates image segmentation and detection techniques to extract surgical knowledge, including vascular maps and guidewire positions. We introduce BDA-star, a novel path planning algorithm with boundary distance constraints, to optimize trajectory planning for guidewire navigation. To validate the method, we developed the KVD-Reinforcement Learning environment, where observations consist of real-time guidewire feeding images highlighting the guidewire tip position and the planned path. We proposed a reward function based on the distances from both the guidewire tip to the planned path and the target to evaluate the agent's actions.Additionally, to address stability issues and slow convergence rates associated with direct learning from raw pixels, we incorporated a pre-trained convolutional neural network into the policy network for feature extraction. Experiments conducted on the aortic simulation autonomous guidewire navigation platform demonstrated that the proposed method, targeting the left subclavian artery, left carotid artery and the brachiocephalic artery, achieved a 100\% guidewire navigation success rate, along with reduced movement and retraction distances and trajectories tend to the center of the vessels.

cs.RO↗