The trained walker actually walks

.zenodo.json

@@ -11,7 +11,7 @@
       "affiliation": "Independent researcher"
     }
   ],
-  "description": "<p><strong>webgpu-fly</strong> runs a whole-animal <em>Drosophila</em> nervous system inside a web browser with no installation and no server. The FlyWire FAFB whole-brain connectome (139,255 neurons, ~15 million synaptic connections) and the Janelia MANC ventral-nerve-cord connectome (23,188 neurons, 5.2 million connections) are each simulated as leaky integrate-and-fire (LIF) networks in fused WebGPU compute kernels &mdash; gather, integrate, threshold and reset in a single kernel, with presynaptic-neurotransmitter signs pre-baked into the connection weights so the inner loop never branches on excitatory/inhibitory type.</p><p>The brain's descending command neurons drive the spinal cord by cell-type name match (the same named cell on both sides of the brain&ndash;VNC boundary), and the spine's 369 leg motor neurons are averaged into a walking magnitude and a turn bias that scale a hand-written tripod gait, which in turn actuates a physically simulated 67-body, 111-actuator <strong>TuragaLab flybody</strong> model running in MuJoCo compiled to WebAssembly. The connectome scales that gait; it does not generate the stepping rhythm, which is an analytic sinusoid of simulation time. A 64&times;16 retina rendered each frame from the fly's own head pose is fed back into the brain's optic neurons, closing a sensorimotor loop. An optional trained reinforcement-learning walking policy (Vaxenburg et al. 2025) runs as a pure-TypeScript forward pass checked element-wise against an independent NumPy re-run of the same extracted weights; that check validates the port's arithmetic, not the assumed layer architecture against the original SavedModel.</p><p>The deployment is a game: the player fires real descending neurons with keypresses to steer the fly to a target, and a winning run produces a deterministic, shareable replay URL that re-executes the identical neuron cascade against the same connectome &mdash; a brain trace, not a video. Performance is reported honestly: the brain LIF kernel is memory-bandwidth-bound and runs at ~0.25 kHz of biological time on an Apple M2 Pro, benchmarked on the same machine against NEST 3.10 (0.67 kHz) and a hand-written multicore Rust port (0.45 kHz). The original 1 kHz target was unreachable for any of the three on that hardware; the contribution is reachability &mdash; a real connectome simulation behind a single URL &mdash; not raw throughput. Known limitations (RL-walker speed gap, closed-loop visual-reflex approximation, kinematic-assist options) are enumerated in LIMITATIONS.md.",
+  "description": "<p><strong>webgpu-fly</strong> runs a whole-animal <em>Drosophila</em> nervous system inside a web browser with no installation and no server. The FlyWire FAFB whole-brain connectome (139,255 neurons, ~15 million synaptic connections) and the Janelia MANC ventral-nerve-cord connectome (23,188 neurons, 5.2 million connections) are each simulated as leaky integrate-and-fire (LIF) networks in fused WebGPU compute kernels &mdash; gather, integrate, threshold and reset in a single kernel, with presynaptic-neurotransmitter signs pre-baked into the connection weights so the inner loop never branches on excitatory/inhibitory type.</p><p>The brain's descending command neurons drive the spinal cord by cell-type name match (the same named cell on both sides of the brain&ndash;VNC boundary), and the spine's 369 leg motor neurons are averaged into a walking magnitude and a turn bias that scale a hand-written tripod gait, which in turn actuates a physically simulated 67-body, 111-actuator <strong>TuragaLab flybody</strong> model running in MuJoCo compiled to WebAssembly. The connectome scales that gait; it does not generate the stepping rhythm, which is an analytic sinusoid of simulation time. A 64&times;16 retina rendered each frame from the fly's own head pose is fed back into the brain's optic neurons, closing a sensorimotor loop. An optional trained reinforcement-learning walking policy (Vaxenburg et al. 2025) runs as a pure-TypeScript forward pass and walks the body from leg actuation and ground reaction alone, with the kinematic assist switched off: 2.004&ndash;2.021 cm per simulated second against a 2.0 cm/s command, uprightness +0.997, no capsize across three repetitions, versus 0.032 cm per simulated second with the policy disabled. That path bypasses the brain and the ventral nerve cord entirely &mdash; it is the published policy walking the fly, not the connectome. The forward pass is checked element-wise against an independent NumPy re-run of the same extracted weights; that check validates the port's arithmetic, not the assumed layer architecture against the original SavedModel.</p><p>The deployment is a game: the player fires real descending neurons with keypresses to steer the fly to a target, and a winning run produces a deterministic, shareable replay URL that re-executes the identical neuron cascade against the same connectome &mdash; a brain trace, not a video. Performance is reported honestly: the brain LIF kernel is memory-bandwidth-bound and runs at ~0.25 kHz of biological time on an Apple M2 Pro, benchmarked on the same machine against NEST 3.10 (0.67 kHz) and a hand-written multicore Rust port (0.45 kHz). The original 1 kHz target was unreachable for any of the three on that hardware; the contribution is reachability &mdash; a real connectome simulation behind a single URL &mdash; not raw throughput. Known limitations (the connectome scaling rather than generating the gait, the closed-loop visual-reflex approximation, kinematic-assist options, the unverified policy architecture) are enumerated in LIMITATIONS.md.",
   "keywords": [
     "WebGPU",
     "WebAssembly",

CITATION.cff

@@ -41,10 +41,17 @@ abstract: >-
   generate its stepping rhythm. A 64x16 retina
   rendered from the fly's head pose feeds back into the brain's optic
   neurons. An optional trained reinforcement-learning walking policy
-  (Vaxenburg et al. 2025) runs as a pure-TypeScript forward pass checked
-  element-wise against an independent NumPy re-run of the same extracted
-  weights, which validates the port's arithmetic but not the assumed
-  layer architecture against the original SavedModel. The project is a game
+  (Vaxenburg et al. 2025) runs as a pure-TypeScript forward pass and walks
+  the body from leg actuation and ground reaction alone, with the
+  kinematic assist off: 2.004-2.021 cm per simulated second against a
+  2.0 cm/s command, uprightness +0.997, no capsize across three
+  repetitions, versus 0.032 cm per simulated second with the policy
+  disabled. That path bypasses the brain and the ventral nerve cord — it
+  is the published policy walking the fly, not the connectome. The
+  forward pass is checked element-wise against an independent NumPy
+  re-run of the same extracted weights, which validates the port's
+  arithmetic but not the assumed layer architecture against the original
+  SavedModel. The project is a game
   with shareable, deterministic replay URLs: a shared link re-executes
   the identical neuron cascade against the same connectome. The brain LIF
   kernel is memory-bandwidth-bound and runs at ~0.25 kHz biological time