Model Predictive Control For Multiple Castaway Tracking with an Autonomous Aerial Agent

TL;DR

This work tackles the problem of tracking multiple castaways at sea with a single autonomous UAV by formulating an online Model Predictive Control (MPC) framework as a nonlinear mixed-integer program (NMIP) that minimizes the aggregate target-state uncertainty. Each castaway is tracked with an individual Kalman Filter, and sensing is modeled as a probabilistic, altitude-dependent observation from the UAV's onboard camera, whose detection probability is learned from CNN-based buoy detectors. The approach integrates a CNN-derived detection probability into the NMIP, planning a rolling-horizon control sequence that respects 3D UAV dynamics, field-of-view constraints, and sensing limitations. Simulation results demonstrate real-time feasibility and reveal trade-offs between horizon length, target count, and computation time, while contributing an open-source buoy-detection dataset to support detector development and future multi-UAV extensions.

Abstract

Over the past few years, a plethora of advancements in Unmanned Areal Vehicle (UAV) technology has paved the way for UAV-based search and rescue operations with transformative impact to the outcome of critical life-saving missions. This paper dives into the challenging task of multiple castaway tracking using an autonomous UAV agent. Leveraging on the computing power of the modern embedded devices, we propose a Model Predictive Control (MPC) framework for tracking multiple castaways assumed to drift afloat in the aftermath of a maritime accident. We consider a stationary radar sensor that is responsible for signaling the search mission by providing noisy measurements of each castaway's initial state. The UAV agent aims at detecting and tracking the moving targets with its equipped onboard camera sensor that has limited sensing range. In this work, we also experimentally determine the probability of target detection from real-world data by training and evaluating various Convolutional Neural Networks (CNNs). Extensive qualitative and quantitative evaluations demonstrate the performance of the proposed approach.

Figures (8)

• Figure 1: Multiple castaway track scenario illustration at two time steps
• Figure 2: Piece wise function emulating the probability of receiving a measurement from the sensor. Parameters were set to $\alpha_1=10, \alpha_2=100$ and $\beta_1=-0.0083, \beta_2=1.083$
• Figure 3: Drift paths of four castaways induced by small amplitude waves for the span of an hour.
• Figure 4: Recall of each CNN at every altitude.
• Figure 5: Sample images of detected buoys at 30 meters altitude. Image (a) shows a successfully detected buoy. Image (b) shows a false positive detection in the absence of buoys. Image (c) shows a false negative detection in the presence of a buoy.