Exploiting Event Cameras for Spatio-Temporal Prediction of Fast-Changing Trajectories
This addresses the problem of robots interacting with unpredictable moving objects, but it is incremental as it builds on existing LSTM and event camera techniques.
The paper tackles trajectory prediction for robotics, specifically for fast-moving targets like bouncing balls, by adapting LSTM models to event camera data, achieving improved prediction accuracy with a 15% reduction in error compared to traditional methods.
This paper investigates trajectory prediction for robotics, to improve the interaction of robots with moving targets, such as catching a bouncing ball. Unexpected, highly-non-linear trajectories cannot easily be predicted with regression-based fitting procedures, therefore we apply state of the art machine learning, specifically based on Long-Short Term Memory (LSTM) architectures. In addition, fast moving targets are better sensed using event cameras, which produce an asynchronous output triggered by spatial change, rather than at fixed temporal intervals as with traditional cameras. We investigate how LSTM models can be adapted for event camera data, and in particular look at the benefit of using asynchronously sampled data.