Training Trajectory Predictors Without Ground-Truth Data

TL;DR

This paper introduces a ground-truth-free framework for estimating vehicle position, heading, and velocity, which enhances trajectory prediction accuracy and robustness without relying on extensive labeled data.

Contribution

The authors develop a novel estimation system that provides high-quality inputs for trajectory prediction models without ground-truth data, improving robustness and applicability in real-world scenarios.

Findings

Poor input quality leads to noisy predictions.

The system enables effective training with limited data.

Robust predictions are achieved across different environments.

Abstract

This paper presents a framework capable of accurately and smoothly estimating position, heading, and velocity. Using this high-quality input, we propose a system based on Trajectron++, able to consistently generate precise trajectory predictions. Unlike conventional models that require ground-truth data for training, our approach eliminates this dependency. Our analysis demonstrates that poor quality input leads to noisy and unreliable predictions, which can be detrimental to navigation modules. We evaluate both input data quality and model output to illustrate the impact of input noise. Furthermore, we show that our estimation system enables effective training of trajectory prediction models even with limited data, producing robust predictions across different environments. Accurate estimations are crucial for deploying trajectory prediction models in real-world scenarios, and our system ensures meaningful and reliable results across various application contexts.