DESCENT: Directed Edge Scene Encoding for Airport Surface Movement Prediction

Noosaga Placements

• Attention Mechanisms and Transformersframework90%
  The architecture is based on transformers, including self-attention and cross-attention mechanisms.
  transformer-based architectureself-attention–based motion encoderattention-based scene encoder
• Roboticssubfield90%
  The paper focuses on trajectory prediction for airport surface movements, a robotics application involving autonomous agents and navigation.
  trajectory prediction architecture tailored to the unique characteristics of airport surface movementsmotion forecasting can be used to verify runway clearance during an aircraft’s final approach
• Deep Learningframework85%
  The model is a deep learning approach for trajectory prediction.
  deep learning-based trajectory predictiontransformer-based architecture
• Deep Learningsubfield85%
  The method is a transformer-based deep learning architecture for trajectory prediction.
  transformer-based architectureattention-based scene encoderdetection transformer-based decoder
• Supervised Deep Learningframework70%
  The model is trained with supervised learning on labeled trajectory data.
  winner-takes-all objectiveNegative Log-Likelihood (NLL) loss
• Machine Learningsubfield60%
  The approach involves supervised learning of trajectory prediction models with loss functions and training on large datasets.
  winner-takes-all objectiveNegative Log-Likelihood (NLL) losstraining on the Amelia-10 benchmark
• Robot Motion Planningsubfield50%
  The paper addresses motion prediction, which is closely related to motion planning, though it does not directly plan trajectories.
  trajectory predictionfuture path of aircraft and ground vehicles

Abstract

Advanced automation is a key technology for enhancing the safety of ground operations amidst the increasing density of commercial air traffic. While motion forecasting is a well-studied task in autonomous driving, its application to airport surface movements remains underexplored. To enable efficient and accurate prediction in this domain, we propose DESCENT, a transformer-based architecture designed to handle heterogeneous dynamics and strict topological constraints. Our approach features a Potential Reachable Set (PRS) context sampling mechanism that adaptively collects airfield environment context across diverse operational phases. Combined with a detection transformer-based decoder, DESCENT generates accurate trajectory forecasts. Extensive evaluations on the Amelia-10 benchmark demonstrate significant performance improvements over state-of-the-art baselines. These gains are especially pronounced in safety-critical scenarios, where our domain-aware sampling provides critical long-horizon context necessary for safe navigation.

Paper Context