Intention-Aware Planner for Robust and Safe Aerial Tracking

Qiuyu Ren, Huan Yu, Jiajun Dai, Zhi Zheng, Jun Meng, Li Xu, Chao Xu, Fei Gao, Yanjun Cao

TL;DR

This paper introduces an "intention-aware planner" for aerial drones that track moving targets. By predicting the target's intentions (like turning or decelerating), the drone can plan safer and more robust trajectories, avoiding occlusions and potential collisions, outperforming existing methods in challenging scenarios.

What Problem Does This Paper Address?

Autonomous aerial tracking, used in applications like filming or surveillance, relies heavily on predicting a target's future movements. However, current tracking systems often assume simple motion models (like constant velocity or rotation), which fail when targets make sudden, aggressive maneuvers (e.g., sharp turns or unexpected stops). This can lead to the drone losing sight of the target (occlusion) or even colliding with obstacles. The core problem this paper addresses is how to make aerial trackers more robust and safer by incorporating an understanding of the target's intentions rather than just its current motion.

Methods

The proposed solution is an "intention-aware planner" with three main components:

  1. Target Intention Prediction: This module uses a novel approach that combines a "user-defined potential assessment function" with a "state observation function." It also generates a "reachable region" to specifically evaluate turning intentions. This allows the system to infer what the target might do next based on its current state and surroundings.
  2. Intention-Driven Target Motion Prediction: An "intention-driven hybrid A*" algorithm is used to predict the target's future positions. This prediction is guided by "intention primitives," which are motion models tailored to specific predicted intentions (e.g., a "turning primitive" or a "deceleration primitive"). This helps generate more realistic future path predictions.
  3. Intention-Aware Trajectory Optimization: The predicted intentions are then integrated into the trajectory optimization process. This involves designing specific "intention-aware constraints" that ensure the drone's planned path maintains visibility and a safe distance from the target, especially when the target is predicted to perform aggressive maneuvers.

Datasets

Not discussed in the paper. The experiments rely on simulated environments and real-world tests with a custom-built quadrotor platform.

Key Insights and Ideas

  • Intentions Matter: The most significant insight is that explicitly modeling and predicting a target's intentions (not just its kinematics) dramatically improves tracking robustness and safety.
  • Reachable Regions for Intent: Using a "reachable region" to assess turning potential is an intuitive way to link geometric information with predicted intentions.
  • Intention Primitives for Prediction: Breaking down motion prediction into "intention primitives" allows for more nuanced and accurate forecasting of future target positions.
  • Coupling Intention with Constraints: Integrating intention prediction directly into trajectory optimization constraints (visibility and safety distance) is crucial for translating inferred intentions into actionable drone behavior.

Key Results

  • Reduced Occlusion: In simulated sharp-turn scenarios, the proposed method significantly reduced occlusion time compared to a state-of-the-art baseline (Ji et al. [5]).
  • Improved Safety: During deceleration scenarios, the intention-aware tracker maintained a more consistent and safer distance from the target, avoiding potential collisions.
  • Higher Success Rate: In real-world turning tests, the intention-aware method achieved an 80% success rate, compared to 30% for the baseline.
  • Robustness: The method demonstrated robustness even at high target velocities.

Limitations and Future Work

  • Single Tracker Limitation: The authors acknowledge that in extremely maneuverable situations, even their advanced tracker might struggle.
  • Future Work: The paper suggests extending this work to "collaborative tracking by a swarm of UAVs," implying that multiple drones working together could further enhance tracking capabilities.

Summary updated

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 speci cally 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.
BibTeX
@inproceedings{ren2024intention,
  author = {Qiuyu Ren and Huan Yu and Jiajun Dai and Zhi Zheng and Jun Meng and Li Xu and Chao Xu and Fei Gao and Yanjun Cao},
  title = {{Intention-Aware Planner for Robust and Safe Aerial Tracking}},
  booktitle = {IEEE/RSJ International Conference on Intelligent Robots and Systems},
  year = {2024},
  url = {https://openreview.net/forum?id=12VOUenzCv},
}

Loading related content...

Figure