RESEARCH NOTE 002 · 2026-09-26
From Navigation to Physical Handoff: Why Arrival Is Not Readiness
Reaching a navigation goal does not mean a mobile robot is ready to receive a tray. In a robot–conveyor handoff, alignment, support, stopped state, receiving capacity, and resource ownership must hold together. This has become a central systems question in our heterogeneous workcell research.
Arrival describes two different tasks
Navigation asks whether the robot can follow a feasible path and satisfy its configured arrival conditions. Physical transfer asks whether the devices are aligned, the gap supports the tray, the receiver has capacity, and both devices have actually stopped.
Navigation success is a result from the navigation subsystem. Transfer permission requires evidence from the handoff subsystem. They can share information, but one result does not establish the other.
Subsystem experiments expose the missing interface
Existing 010 records include single-robot arrival, conveyor handoffs, and subsystem experiments in a shared simulation world. Arrival repeatability and docking tolerance both require independent checks. A transfer from a prealigned starting position does not establish autonomous transfer after navigation.
We do not compare error numbers across those experiments here: world versions, frames, payloads, and control conditions may differ. The records identify an interface gap; they do not establish stable autonomous handoff.
Make the stages between navigation and transfer explicit
A workflow to validate is: navigate to a waiting area → observe relative docking pose → adjust at low speed → verify actual stopping → check receiving capacity and resource permission → transfer → verify reception and source clearance. Each stage needs inputs, exit conditions, deadlines, and failure reasons.
Stale observations, excessive alignment error, an occupied receiver, or an unknown stopped state must keep transfer blocked. A stop command does not prove physical stopping, and one receiver sensor does not prove a complete transfer.
Precision docking also depends on interface design
If transfer succeeds only within a very small geometric range, localization tuning alone may not provide enough margin. Mechanical guides, continuous support, gap size, conveyor heights, and the ability to measure relative pose also need investigation.
Measure the allowed geometry first, then the distribution of control errors, and finally the remaining margin. Do not derive acceptance thresholds from one successful run or relax them to produce a passing report.
Separate control feedback from acceptance evidence
Controllers should use the permitted sensor observations. Simulation truth can support independent evaluation, but must not silently become control feedback. Reports should retain software and world versions, configuration, task outcomes, physical outcomes, and missing checks.
The next useful loop is single-robot approach, docking, tray reception, loaded transport, destination docking, and unloading. Arm loading, humanoid tray supply, and two-robot coordination can follow. The complete loop is still under integration research.
Reliability begins with clear boundaries
Integration failures often occur between two subsystems that appear successful on their own. Distinguishing navigation complete, docking verified, transfer permitted, and payload received makes failures explainable and ownership consistent, while giving improvements a clear validation target.
This article records the current engineering judgment and proposed validation method. It does not claim hardware deployment, industrial safety certification, or complete workcell acceptance.