Loading…
Loading…
k9 is my physical robot prototype for connecting software control with legged hardware. The gallery shows the assembled robot, its wiring and short behavior clips. It documents a build and demonstration, not a measured autonomous-navigation result.
The assembly photo shows a four-legged metal frame, servo-driven joints, circuit boards and exposed wiring. The behavior clip shows the robot on a floor with a front-mounted sensor assembly. The photographs establish the physical arrangement; they do not identify the controller firmware or show how commands are chosen.
The recorded clips show the prototype operating during selected demonstrations. They do not show the operator input, a complete perception-to-action trace, a repeated task protocol or a baseline controller. I therefore do not infer autonomous decision-making from movement alone.
The original goal was to connect AI software with hardware. The current evidence does not yet separate AI-generated commands, a scripted sequence and direct remote control. That distinction matters: a remote command can demonstrate actuator control without demonstrating an autonomous policy.
I assembled and demonstrated the prototype. The exact board models, kit origin, firmware changes, model integration and which control software I wrote still need a source-backed build record. I do not claim that I designed the servo hardware, trained a locomotion policy or implemented autonomous navigation.
A useful next hardware test is a fixed command sequence on the same floor: start from a marked pose, execute the requested motion, return to a safe idle state, and stop on command. Record the operator input and the robot in one uninterrupted clip. Compare manual or scripted control with the AI-assisted version only after the control path is documented.
For each attempt, record the initial pose, command, controller mode, completion criterion, elapsed time and failure reason. Keep unsuccessful attempts. This makes a completed task distinguishable from an edited highlight.
No task success rate, collision rate, command latency, power budget or comparison with a baseline is reported here. The existing gallery also does not establish recovery from a fall or a lost connection. A complete task and failure recording, plus the control architecture, are the remaining evidence needed for those claims.
For now, read k9 as a hardware integration prototype with recorded behavior. The clips and assembly photos below are the evidence available today.