Robot Trajectron V2: A Probabilistic Shared Control Framework for Navigation

2025-09-24

AI Summary

This work addresses the dual challenges of inaccurate user-intent prediction and the trade-off between safety and autonomy in shared-control navigation. We propose a probabilistic shared-control framework that jointly integrates a prior intent model with a posterior Bayesian updating mechanism. Innovatively, it unifies multimodal inputs (e.g., keyboard, brain–computer interface), temporally dependent behavioral history, and real-time environmental context via a hybrid architecture combining recurrent neural networks (RNNs) and conditional variational autoencoders (CVAEs). The framework further incorporates reinforcement learning for adaptive control policy optimization and safety-constrained trajectory planning. Evaluated on synthetic benchmarks, human-subject experiments, and non-human primate BCI deployments, our method achieves statistically significant improvements in intent estimation accuracy while better balancing user autonomy and system intervention. It delivers safe, robust, and scalable navigation support across diverse assistive technology scenarios.

Technology Category

• Intelligent Robots: Behavior Learning & Control
• Humans and AI: Human-Aware Planning and Behavior Prediction
• Cognitive Modeling & Cognitive Systems: Adaptive Behavior

Application Category

• Responsible Web: Machine-in-the-loop, human agency and autonomy
• Economics, Online Markets and Human Computation: The sharing economy
• User Modeling, Personalization and Recommendation: Psychology-informed user models and recommender systems

Abstract

We propose a probabilistic shared-control solution for navigation, called Robot Trajectron V2 (RT-V2), that enables accurate intent prediction and safe, effective assistance in human-robot interaction. RT-V2 jointly models a user's long-term behavioral patterns and their noisy, low-dimensional control signals by combining a prior intent model with a posterior update that accounts for real-time user input and environmental context. The prior captures the multimodal and history-dependent nature of user intent using recurrent neural networks and conditional variational autoencoders, while the posterior integrates this with uncertain user commands to infer desired actions. We conduct extensive experiments to validate RT-V2 across synthetic benchmarks, human-computer interaction studies with keyboard input, and brain-machine interface experiments with non-human primates. Results show that RT-V2 outperforms the state of the art in intent estimation, provides safe and efficient navigation support, and adequately balances user autonomy with assistive intervention. By unifying probabilistic modeling, reinforcement learning, and safe optimization, RT-V2 offers a principled and generalizable approach to shared control for diverse assistive technologies.

Problem

• Develops probabilistic shared control for human-robot navigation
• Models user intent and noisy control signals jointly
• Enables safe assistance while balancing user autonomy

Innovation

• Probabilistic shared control combining intent prediction and assistance
• Joint modeling of behavioral patterns with real-time user inputs
• RNN and variational autoencoder framework for multimodal intent inference