📄 中文摘要
本研究聚焦于在复杂环境中具有旅行预算的信息路径规划（IPP）问题。智能体通过收集对高斯过程（GP）建模的潜在场的测量数据，以降低目标位置的不确定性。现有解决方案中，基于图的方法能够提供全局保证，但前提是预先选择测量位置。然而，连续轨迹优化方法虽然支持基于路径的传感，但计算成本高昂，且对初始化敏感。为了弥补这些不足，本研究提出了一种分层信息路径规划框架。该框架首先利用图引导策略进行粗粒度路径规划，通过构建环境的稀疏图表示，在全局范围内识别潜在的有效测量区域。图引导阶段的目标是在满足旅行预算的前提下，最大化信息增益，同时避免陷入局部最优。随后，在图引导生成的粗略路径基础上，引入轨迹优化技术进行精细化调整。轨迹优化阶段着重于平滑路径，优化传感器的姿态和扫描策略，以进一步提升信息收集效率和路径平滑度。这种分层方法结合了图方法的全局搜索能力和轨迹优化的局部精细调整能力，旨在克服传统方法的局限性。具体而言，图引导步骤有助于在广阔的搜索空间中快速定位有前景的区域，显著降低了后续轨迹优化的初始化敏感性。轨迹优化则确保了生成的路径在实际部署中的可行性和高效性，例如考虑了动态障碍物规避、传感器视场约束以及机器人动力学限制。通过这种结合，该方法能够在保持计算效率的同时，实现高质量的信息收集，尤其适用于资源受限且环境复杂的应用场景，例如环境监测、灾害评估和自主探索等。
📄 English Summary
Hierarchical Informative Path Planning via Graph Guidance and Trajectory Optimization
This research investigates informative path planning (IPP) with travel budgets in cluttered environments, where an agent collects measurements of a latent field modeled as a Gaussian process (GP) to reduce uncertainty at target locations. Graph-based solvers offer global guarantees but rely on pre-selected measurement locations. Conversely, continuous trajectory optimization supports path-based sensing but is computationally intensive and sensitive to initialization. To address these limitations, a hierarchical informative path planning framework is proposed. This framework initially employs a graph-guided strategy for coarse-grained path planning. It constructs a sparse graph representation of the environment to identify globally promising measurement regions within the travel budget, aiming to maximize information gain while avoiding local optima. Subsequently, building upon the rough paths generated by the graph guidance, trajectory optimization techniques are introduced for fine-grained refinement. The trajectory optimization phase focuses on smoothing the path and optimizing sensor poses and scanning strategies to further enhance information collection efficiency and path smoothness. This hierarchical approach combines the global search capabilities of graph methods with the local fine-tuning abilities of trajectory optimization, aiming to overcome the limitations of traditional methods. Specifically, the graph-guided step helps quickly locate promising areas in a vast search space, significantly reducing the initialization sensitivity of subsequent trajectory optimization. Trajectory optimization then ensures the feasibility and efficiency of the generated paths in practical deployment, considering factors such as dynamic obstacle avoidance, sensor field-of-view constraints, and robot dynamics. Through this combination, the method achieves high-quality information gathering while maintaining computational efficiency, making it particularly suitable for resource-constrained and complex environmental applications, such as environmental monitoring, disaster assessment, and autonomous exploration.