# BRI-002: Autonomy Infused Teleoperation with Application to BCI Manipulation

> 本页由本地论文卡片生成。请不要直接编辑本页；修改源卡片后运行 `vp run docs:generate`。

> 中文支持说明：本页是中文站点镜像，保留论文标题、证据状态和来源字段的原始表述，避免未经核实的翻译改变证据边界。

> 内部 wiki 边界：本页提供论文卡片证据和用户提供的 PDF 下载入口，仅用于私有阅读；全文抽取文本不发布。

> 来源: `library/paper_cards/BRI-002.md`

## 论文访问

* 内部 PDF: <a href={"/papers/BRI-002.pdf"} download style={{ display: "inline-flex", alignItems: "center", justifyContent: "center", minHeight: "2.25rem", padding: "0.45rem 0.8rem", borderRadius: "6px", backgroundColor: "#047857", color: "#ffffff", fontWeight: 700, lineHeight: 1, textDecoration: "none", boxShadow: "0 1px 2px rgba(15, 23, 42, 0.22)" }}>下载论文 PDF</a>
* DOI / 官方页面: [needs confirmation](https://arxiv.org/abs/1503.05451)
* 开放访问页面: [Open access page](https://arxiv.org/abs/1503.05451)
* 部署边界: 这些 PDF 链接只适合私有/受保护的内部 wiki。

## BRI-002: Autonomy Infused Teleoperation with Application to BCI Manipulation

## Metadata

* ID: BRI-002
* Title: Autonomy Infused Teleoperation with Application to BCI Manipulation
* Year: 2015
* DOI / URL: https://arxiv.org/abs/1503.05451
* Local PDF: 见上方论文访问区块
* Text artifact: local-only path withheld from docs site
* Review status: `extracted`

## Study Type

* Track: Hybrid BCI / Brain-Robot Interface
* Task: autonomy-infused teleoperation for BCI robotic manipulation
* Participants or dataset: 2 participants with tetraplegia using intracortical BCI; Subject 1 had 2.5 years BCI use, Subject 2 had 2 months BCI use
* Hardware: RGB-D camera on a 2-stage 4-axis neck; 7-DOF Barrett WAM; wrist force-torque sensor; 4-DOF BarrettHand with pressure sensors
* Channels or sensors: Blackrock intracortical microelectrode arrays; Subject 1 had two arrays, Subject 2 had four arrays

## Methods

* Paradigm: adapted ARAT, Box and Blocks, multi-object grasping, and manipulation demonstrations comparing autonomy-infused teleoperation with direct control
* Signal processing or model: threshold-crossing neural features; indirect OLE decoder for endpoint/grasp velocities; AIT with depth-template object perception, capture envelopes for grasp inference, MaxEnt goal inference, linear arbitration, and compliant/safety control
* Training/calibration: intracortical BCI decoder calibration and task practice; Subject 1 had more prior practice than Subject 2
* Online/offline: online BCI shared-control robotic manipulation

## Results

* Metrics: in adapted ARAT, Subject 1 achieved 13/15 complete successes with AIT and 0/15 with direct control; Subject 2 achieved 13/15 grasps with AIT and 0/15 with direct control; Box and Blocks transfers averaged 5.2 with AIT versus 2.6 with direct control; Subject 2 selected the specified block 32/36 times in multi-object grasping
* Main findings: computer vision, intent inference, and arbitration substantially improved or enabled BCI manipulation relative to direct low-level control in these tasks
* Reported limitations: intracortical BCI is low-dimensional and noisy; long-term invasive signal degradation is a risk; object models and grasp poses were predefined; novel-object extension was only proof-of-concept with a game controller

## Relevance To This Project

* Supports: the core architecture of BCI intent plus vision-based shared autonomy for robotic grasping
* Conflicts with: the paper uses intracortical BCI rather than noninvasive EEG/SSVEP-MI
* Design implication: noninvasive EEG should probably provide higher-level, lower-rate intent while robot vision and shared autonomy handle grasp inference and execution

## Extracted Evidence

| Claim | Status | Evidence Note | Page/Section |
| --- | --- | --- | --- |
| Vision-assisted autonomy can improve BCI robotic manipulation compared with direct BCI control. | verified | AIT improved adapted ARAT and Box and Blocks outcomes relative to direct control in the reported participants. | ARAT; Object Transfer |
| Intent inference and arbitration are central mechanisms for BCI shared-control manipulation. | verified | The AIT architecture combines object perception, MaxEnt goal inference, and linear arbitration between autonomy and user input. | Autonomous Robot Manipulation Assistance |
| Direct low-level BCI control is limited by noisy and low-dimensional input. | verified | The paper describes BCI control as low-dimensional/noisy and reports direct-control failures in the manipulation tasks. | Experimental setup; Results; Limitations |
| SAH-BRI-Grasp should use noninvasive EEG as a high-level intent source and delegate grasp execution to vision/shared autonomy. | inferred | The paper supports BCI+vision+autonomy, but its BCI is intracortical; noninvasive EEG transfer requires additional evidence and experiments. | Conclusion |

## Open Questions

* Which AIT components can be simplified for a noninvasive EEG prototype?
* How should YOLO object candidates replace predefined object models?
* Can shared autonomy remain usable when SSVEP/MI commands are slower and more discrete than intracortical velocity control?