# Open Questions

> Generated from repository source files. Do not edit this page directly; edit the source file listed below and rerun `vp run docs:generate`.

> Source: `research/05-open-questions.md`

## Open Questions

Status: `wiki planning source`

This page lists the questions that should drive reading, experiments, manuscript writing, and product exploration.

## P0 Questions

| Question                                                                           | Current Status       | Evidence Now                                                                      | What Would Resolve It                                                        |
| ---------------------------------------------------------------------------------- | -------------------- | --------------------------------------------------------------------------------- | ---------------------------------------------------------------------------- |
| Can dynamic scene objects function as SSVEP targets?                               | `needs confirmation` | SSVEP fixed-layout, benchmark, AR/VR, and decoder evidence exists.                | Exp1 comparing fixed, frozen-scene, and live-scene candidates.               |
| Can MI work as mode control without unsafe false activation?                       | `needs confirmation` | MI classifier, no-control benchmark context, and robot-control comparators exist. | Exp2 with rest/no-control, latency, false activation, and rejection metrics. |
| Does scene-aware shared autonomy improve closed-loop grasping over fixed commands? | `needs confirmation` | Shared autonomy and BRI manipulation comparators exist.                           | Exp3 with fixed-command, scene-aware, single-modality, and hybrid baselines. |
| Can the system be reduced to fewer EEG channels without losing feasibility?        | `needs confirmation` | SSVEP/MI low-channel and compact decoder evidence exists.                         | Exp4 offline ablation and then online reduced-channel pilot.                 |
| Can YOLO outputs become safe grasp commands?                                       | `needs confirmation` | Object detection and grasping/calibration evidence exists.                        | Project grasp-pose implementation, calibration records, and Exp3 logs.       |

## Design Questions

| Question                                   | Working Answer                                                                                | Status                                                             |
| ------------------------------------------ | --------------------------------------------------------------------------------------------- | ------------------------------------------------------------------ |
| Should EEG control robot joints directly?  | No; non-invasive EEG should express high-level intention and intervention.                    | `inferred`                                                         |
| What is SSVEP responsible for?             | Target selection, confirmation, cancel, or other discrete visual choices.                     | `inferred`                                                         |
| What is MI responsible for?                | Mode control, active intervention, stop, or confirmation, not dense 6-DoF control.            | `inferred`                                                         |
| What does YOLO contribute?                 | It generates a dynamic command space from physical scene objects.                             | `inferred`                                                         |
| What does shared autonomy contribute?      | It maps uncertain high-level user intent to feasible robot action.                            | `inferred`                                                         |
| What is the current contribution boundary? | A scene-aware hybrid BRI framework and grasping system design; performance is not proven yet. | `verified` for project scope; `needs confirmation` for performance |

## Literature Questions

| Track                  | Question To Ask While Reading                                       | Current Use                                      |
| ---------------------- | ------------------------------------------------------------------- | ------------------------------------------------ |
| BCI foundations        | What limits should constrain non-invasive EEG claims?               | Writing boundaries and evaluation discipline.    |
| SSVEP                  | What transfers from fixed targets to object-bound scene candidates? | Exp1 design and target-selection rationale.      |
| MI                     | What evidence supports mode-level control and no-control rejection? | Exp2 design and safety boundary.                 |
| YOLO / detection       | What detector failures matter for BCI target binding?               | Candidate generation and Exp1/Exp3 failure logs. |
| Grasping / calibration | Where does detection end and grasp execution begin?                 | Method boundary and Exp3 hardware binding.       |
| Hybrid BRI             | What comparable systems combine BCI, vision, and autonomy?          | Related work and contribution positioning.       |
| Shared autonomy        | How should uncertain intent be blended with robot policy?           | State machine and baseline design.               |

## Product Questions

| Direction                              | Key Question                                                                         | Status                                                            |
| -------------------------------------- | ------------------------------------------------------------------------------------ | ----------------------------------------------------------------- |
| Research paper output suite            | Can this workflow become an evidence-backed manuscript production tool?              | `inferred`; needs user workflow validation.                       |
| Embodied AI + BCI concept              | Can SAH-BRI become a convincing demo of human intention for embodied robot autonomy? | `inferred`; needs working demo evidence.                          |
| Rehabilitation / assistive interaction | Can the system support clinical or assistive value?                                  | `needs confirmation`; requires clinical protocol and safety case. |

## Dexterous Arm-Hand Extension Questions

This is an exploratory direction and is not a new claim about the current SAH-BRI-Grasp implementation.

| Question                                                                                                      | Current Status       | What Would Resolve It                                                                        |
| ------------------------------------------------------------------------------------------------------------- | -------------------- | -------------------------------------------------------------------------------------------- |
| Can SSVEP-MI supervise a high-dimensional arm-hand system through a small grasp-skill library?                | `needs confirmation` | local hand integration followed by online comparison against simpler command interfaces      |
| Does an object-task-skill hierarchy improve task success without adding excessive BCI selections or workload? | `needs confirmation` | staged user study with task success, decision count, completion time, and workload measures  |
| When should the robot choose a grasp skill and when should it request user disambiguation?                    | `needs confirmation` | confidence-aware arbitration experiment                                                      |
| Can an open-source dexterous hand execute repeatable primitive grasps on the selected self-designed arm?      | `needs confirmation` | frozen hardware, object set, thresholds, repeated trials, and failure logs                   |
| Does feedback-aware closure improve safety and reliability over position-only closure?                        | `needs confirmation` | paired hardware trials with contact, current, temperature, and abort records where available |

## Current Blockers

* No Exp1-Exp4 data exist yet.
* Hardware inventory and middleware binding are not complete.
* Hand-eye calibration protocol is not bound to local hardware.
* The dexterous arm-hand direction has no selected hardware revision, local literature lane, skill implementation, or experiment protocol.
* Results, result-dependent Discussion, and Conclusion must stay blocked.
* Clinical and product-market claims remain outside the verified evidence boundary.

## Next Step

Use the [Claim Map](/research-map/claim-map) to inspect candidate claims, or move to the [Experiment Roadmap](/experiments/roadmap) to follow the validation order.