Neural Motion Prediction for In-flight Uneven Object Catching

TL;DR

This paper introduces a neural acceleration estimation method combined with differentiable filtering to improve in-flight uneven object catching, demonstrating high success rates and generalization in real-world robotic experiments.

Contribution

The paper proposes the Neural Acceleration Estimator (NAE) and NAE-DF for accurate motion prediction of uneven objects, enhancing robotic catching performance.

Findings

Achieved over 83% success rate in real-world catching tasks.

Demonstrated superior prediction accuracy over existing methods.

Released a dataset with 1,500 trajectories for uneven objects.

Abstract

In-flight objects capture is extremely challenging. The robot is required to complete trajectory prediction, interception position calculation and motion planning in sequence within tens of milliseconds. As in-flight uneven objects are affected by various kinds of forces, motion prediction is difficult for a time-varying acceleration. In order to compensate the system's non-linearity, we introduce the Neural Acceleration Estimator (NAE) that estimates the varying acceleration by observing a small fragment of previous deflected trajectory. Moreover, end-to-end training with Differantiable Filter (NAE-DF) gives a supervision for measurement uncertainty and further improves the prediction accuracy. Experimental results show that motion prediction with NAE and NAE-DF is superior to other methods and has a good generalization performance on unseen objects. We test our methods on a robot, performing velocity control in real world and respectively achieve 83.3% and 86.7% success rate on a ploy urethane banana and a gourd. We also release an object in-flight dataset containing 1,500 trajectorys for uneven objects.