# End-to-end Optimization of Belief and Policy Learning in Shared Autonomy Paradigms

arXiv: https://arxiv.org/abs/2601.23285 · HTML: https://mlsanity.com/paper/2601.23285/

## Key facts

- No citing papers in the map yet (data through 14 Oct 2026).
- Builds on 19 of its 37 references that are in the map.
- First posted on arXiv 30 Jan 2026.

## Builds on (most cited first, 19 of 19)

- [Proximal Policy Optimization Algorithms](https://mlsanity.com/paper/1707.06347/) (2017)
- [Shared Autonomy via Deep Reinforcement Learning](https://mlsanity.com/paper/1802.01744/) (2018)
- [Shared Autonomy via Hindsight Optimization](https://mlsanity.com/paper/1503.07619/) (2015)
- [Shared Autonomy with Learned Latent Actions](https://mlsanity.com/paper/2005.03210/) (2020)
- [Residual Policy Learning for Shared Autonomy](https://mlsanity.com/paper/2004.05097/) (2020)
- [Blending Imitation and Reinforcement Learning for Robust Policy Improvement](https://mlsanity.com/paper/2310.01737/) (2023)
- [Human-Robot Mutual Adaptation in Shared Autonomy](https://mlsanity.com/paper/1701.07851/) (2017)
- [The Sense of Agency in Assistive Robotics Using Shared Autonomy](https://mlsanity.com/paper/2501.07462/) (2025)
- [SARI: Shared Autonomy across Repeated Interaction](https://mlsanity.com/paper/2205.09795/) (2022)
- [Bayesian Intention for Enhanced Human Robot Collaboration](https://mlsanity.com/paper/2410.00302/) (2024)
- [A Bayesian-Based Approach to Human Operator Intent Recognition in Remote Mobile Robot Navigation](https://mlsanity.com/paper/2109.12045/) (2021)
- [A novel seamless magnetic-based actuating mechanism for end-effector-based robotic rehabilitation platforms](https://mlsanity.com/paper/2404.01441/) (2024)
- [Shared Autonomy with IDA: Interventional Diffusion Assistance](https://mlsanity.com/paper/2409.15317/) (2024)
- [Learning to Arbitrate Human and Robot Control using Disagreement between Sub-Policies](https://mlsanity.com/paper/2108.10634/) (2021)
- [A Comparative Study of Conventional and Tripolar EEG for High-Performance Reach-to-Grasp BCI Systems](https://mlsanity.com/paper/2402.09448/) (2024)
- [STREAMS: An Assistive Multimodal AI Framework for Empowering Biosignal Based Robotic Controls](https://mlsanity.com/paper/2410.03486/) (2024)
- [Learning Multimodal AI Algorithms for Amplifying Limited User Input into High-dimensional Control Space](https://mlsanity.com/paper/2505.11366/) (2025)
- [Human-Centered Shared Autonomy for Motor Planning, Learning, and Control Applications](https://mlsanity.com/paper/2506.16044/) (2025)
- [Macroscopic EEG Reveals Discriminative Low-Frequency Oscillations in Plan-to-Grasp Visuomotor Tasks](https://mlsanity.com/paper/2510.19057/) (2025)

## Abstract (arXiv)

Shared autonomy systems require principled methods for inferring user intent and determining appropriate assistance levels. This is a central challenge in human-robot interaction, where systems must be successful while being mindful of user agency. Previous approaches relied on static blending ratios or separated goal inference from assistance arbitration, leading to suboptimal performance in unstructured environments. We introduce BRACE (Bayesian Reinforcement Assistance with Context Encoding), a novel framework that fine-tunes Bayesian intent inference and context-adaptive assistance through an architecture enabling end-to-end gradient flow between intent inference and assistance arbitration. Our pipeline conditions collaborative control policies on environmental context and complete goal probability distributions. We provide analysis showing (1) optimal assistance levels should decrease with goal uncertainty and increase with environmental constraint severity, and (2) integrating belief information into policy learning yields a quadratic expected regret advantage over sequential approaches. We validated our algorithm against SOTA methods (IDA, DQN) using a three-part evaluation progressively isolating distinct challenges of end-effector control: (1) core human-interaction dynamics in a 2D human-in-the-loop cursor task, (2) non-linear dynamics of a robotic arm, and (3) integrated manipulation under goal ambiguity and environmental constraints. We demonstrate improvements over SOTA, achieving 6.3% higher success rates and 41% increased path efficiency, and 36.3% success rate and 87% path efficiency improvement over unassisted control. Our results confirmed that integrated optimization is most beneficial in complex, goal-ambiguous scenarios, and is generalizable across robotic domains requiring goal-directed assistance, advancing the SOTA for adaptive shared autonomy.