Intention-Aware Planner for Robust and Safe Aerial Tracking

TL;DR

The paper tackles robust aerial tracking under aggressive target maneuvers by addressing the limitation of constant-velocity targets. It introduces an intention-aware planner that couples target intention prediction, intention-driven motion prediction via a hybrid A* with intention primitives, and intention-aware trajectory optimization using MINCO parameterization. Key contributions include a novel intention prediction framework combining a potential-based and state-observation approach, an intention-driven search for future target positions, and a trajectory optimizer that enforces intention-informed visibility and safety constraints. Experimental results in simulation and real-world flights demonstrate improved visibility, reduced occlusions, and safer tracks in cluttered environments, highlighting the practical impact for cinematography, surveillance, and pursuit tasks.

Abstract

Autonomous target tracking with quadrotors has wide applications in many scenarios, such as cinematographic follow-up shooting or suspect chasing. Target motion prediction is necessary when designing the tracking planner. However, the widely used constant velocity or constant rotation assumption can not fully capture the dynamics of the target. The tracker may fail when the target happens to move aggressively, such as sudden turn or deceleration. In this paper, we propose an intention-aware planner by additionally considering the intention of the target to enhance safety and robustness in aerial tracking applications. Firstly, a designated intention prediction method is proposed, which combines a user-defined potential assessment function and a state observation function. A reachable region is generated to specifically evaluate the turning intentions. Then we design an intention-driven hybrid A* method to predict the future possible positions for the target. Finally, an intention-aware optimization approach is designed to generate a spatial-temporal optimal trajectory, allowing the tracker to perceive unexpected situations from the target. Benchmark comparisons and real-world experiments are conducted to validate the performance of our method.

Figures (10)

• Figure 1: Comparisons between our intention-aware tracker and a tracker without considering intentionji2022elastic. As the target passes through a T-shaped intersection, the intention-aware tracker can maintain better visibility without any occlusions compared to the other.
• Figure 2: Overview of the proposed system with the designed drone. The system is supported by a perception module, processing data from sensors on the quadrotor to perceive the environment, shown on the left. The core part of the system, the intention-aware planning module, includes three parts, the target intention prediction part in dark yellow, the target motion prediction part in blue, and the intention-aware trajectory optimization part in green.
• Figure 3: Target localization and pose estimation. Left: Human joint estimation using the Mediapipe framework. Middle: Target localization utilizing four joints of the trunk. Right: Target orientation estimation based on the orientation of its shoulders.
• Figure 4: The reachable region is generated based on the target's motion state and the surrounding environment.
• Figure 5: Target motion prediction and occlusion-free region setup when the target turns right.