In context
As low-altitude economies expand, urban airspace is seeing increasing numbers of heterogeneous, high-density UAV operations, including delivery and air-taxi services. However, monitoring and safeguarding these flights remain challenging, especially against unauthorized drones. Accurate trajectory prediction is critical for collision avoidance and risk warning, but urban environments with frequent maneuvering and obstacles complicate long-term forecasting.
What was reported
Researchers at Harbin Institute of Technology proposed a hierarchical trajectory prediction framework that fuses intent inference with local planning for low-altitude flight targets in urban settings. The method discretizes the urban area into traversable blocks, using an LSTM network to learn a trajectory transition probability model from historical data. This reduces the difficulty of predicting coordinates directly.
Bayesian inference then updates the posterior probability distribution of the target's destination site online, capturing long-term motion intent. Predicted trajectories are generated via sampling and local planning, ensuring compliance with UAV dynamic constraints while reducing computational complexity for real-time performance.
Experiments showed that destination site estimation achieved an average accuracy of 0.46 (mean ± standard error: 0.175) within a range of 9 target blocks, significantly outperforming methods without prior information. The predicted trajectories covered potential maneuver paths with smaller errors and faster switching speeds compared to the interactive multiple models (IMM) algorithm, offering advantages in long-term prediction.
Why it mattered
This work addresses a key gap in urban airspace management by combining deep learning with Bayesian reasoning for intent-aware trajectory prediction. It provides a practical approach for monitoring systems to anticipate UAV behavior, enhancing safety and efficiency in low-altitude operations—a foundation for scalable urban air mobility.
“The predicted trajectories cover potential maneuver paths of the target, exhibiting smaller errors, faster switching speeds, and superior performance in long-term trajectory prediction compared to the interactive multiple models (IMM) algorithm.”
Source: 《机器人》期刊 (robot.sia.cn) · Published 2025-06-09 · “基于意图推理的城市环境下低空飞行目标轨迹预测方法”