An Informative Planning Framework for Target Tracking and Active Mapping in Dynamic Environments with ASVs

TL;DR

This work addresses active mapping and tracking of freely drifting floating targets under wind-driven disturbances using an autonomous surface vehicle. It introduces an integrated pipeline that combines dynamic occupancy grid mapping with a spatiotemporal prediction network and a drift-aware planning objective, enabling efficient re-detection and accurate target localization over time. The key contributions include a wind-driven drift model within dynamic occupancy grids, a UNet-based spatiotemporal predictor that outputs target position distributions, and a planning utility that jointly optimizes exploration and target re-detection using forward-looking predictions. Simulations demonstrate improved target tracking and mapping accuracy over entropy-only baselines, and field tests validate real-world viability, with open-source software available for replication. The approach advances real-time adaptive sensing in dynamic aquatic environments and offers a path toward multi-robot extensions and autonomous drift inference.

Abstract

Mobile robot platforms are increasingly being used to automate information gathering tasks such as environmental monitoring. Efficient target tracking in dynamic environments is critical for applications such as search and rescue and pollutant cleanups. In this letter, we study active mapping of floating targets that drift due to environmental disturbances such as wind and currents. This is a challenging problem as it involves predicting both spatial and temporal variations in the map due to changing conditions. We introduce an integrated framework combining dynamic occupancy grid mapping and an informative planning approach to actively map and track freely drifting targets with an autonomous surface vehicle. A key component of our adaptive planning approach is a spatiotemporal prediction network that predicts target position distributions over time. We further propose a planning objective for target tracking that leverages these predictions. Simulation experiments show that this planning objective improves target tracking performance compared to existing methods that consider only entropy reduction as the planning objective. Finally, we validate our approach in field tests, showcasing its ability to track targets in real-world monitoring scenarios.

Figures (7)

• Figure 1: Our informative path planning (IPP) framework applied to monitoring freely drifting targets (colored buoy markers) with an autonomous surface vehicle that is equipped with a stereo camera. A dynamic mapping strategy is used for target tracking. We introduce a spatiotemporal prediction network and a new IPP objective that enable accurate target tracking.
• Figure 2: An overview of our IPP framework (in blue) for active mapping of dynamic targets. We detect and localize targets with a stereo camera on an ASV. We use a dynamic occupancy grid mapping method that updates the map with both the sensor percepts and predicted motion of the targets. We introduce our spatiotemporal network that enables efficient predictions to aid the adaptive planning approach.
• Figure 3: Localization error $\sigma(r)$ as function of distance to detection $r$ based on empirical analysis (left). The inverse sensor models used for mapping are shown on the right.
• Figure 4: An overview of our spatiotemporal prediction network. The inputs are a filtered grid with only target positions, along with a vector of the wind speeds $v_x$, $v_y$, and prediction interval $t$. The output is the predicted target position distributions at $t$.
• Figure 5: Entropy reduction $\Delta H$ and MSE for planning with the sampling-based planner, with and without the prediction step of the dynamic occupancy grid. With the prediction step, the mapping accuracy is observed to be better by $21.8\%$.