Fusing high-level cognitive understanding with bio-inspired spatial adaptability. Engineered with a Rockchip RV1126B Linux compute core, dual Sony IMX415 vision, expressive GC9A01 circular eye displays, and holonomic mobility.
Harmonizing live duplex Gemini Live voice, Sony IMX415 visual context, acoustic DoA gaze tracking, and adaptive spatial conductivity networks.
Inspired by Physarum polycephalum dynamics. Rather than static graph searches, our custom Rust engine models space as a flux-driven continuous conductivity field for dynamic obstacle bypass and adaptive corridor discovery.
Employs potential difference equations and non-linear conductivity updates: high-demand corridors reinforce while dormant paths decay smoothly over time.
Engineered from scratch in Rust with zero heavy framework bloat. Built for zero-cost abstraction, predictable memory latency, and future cross-compilation onto ARM edge targets.
Leveraging Google Cloud compute clusters for large-scale Monte Carlo parameter sensitivity sweeps, benchmarking candidate channel generation before physical embedded deployment.
Seamless integration of natural conversation, visual perception, and expressive physical embodiment.
Real-time bidirectional speech streams with seamless user barge-in and audio cancellation. Children converse spontaneously with zero awkward waiting.
ReSpeaker XVF3800 AEC + Live WSSShow toys, sketches, or everyday items. The robot grabs live frames via Sony IMX415 sensors, analyzes semantics via Gemini Vision, and spins educational tales.
Dual Sony IMX415 + Gemini Vision RelayA three-wheel omnidirectional chassis (Ø142mm base) driven by high-torque L-type 520 geared motors, engineered for indoor domestic traversal.
Omni-directional Drive + Fall-prevention ToFStrict separation of high-level cloud cognition, edge sensory coordination, and safety-critical motion fail-safes.
Real-time multimodal speech comprehension, visual contextual reasoning, situational storytelling generation, and large-scale spatial simulation orchestration.
Independent motion brake conditions, 3-wheel holonomic kinematics, VL53L0X optical drop detection, ICM-42688-P 6-axis IMU, and RPLIDAR C1 collision guard.
Strictly engineered with verified engineering prototypes.
| Component Subsystem | Hardware Specification | Functional Role |
|---|---|---|
| Main Compute SoC | Rockchip RV1126B (Forlinx OK1126B-S) | Embedded Linux, NPU acceleration, Rust core daemon |
| Vision Sensor | Sony IMX415 (MIPI-CSI, 4K Support) | Dual camera stream, 1280x720 30FPS low-latency feed |
| Audio Frontend | ReSpeaker / XVF3800 Array (USB) | 6-Channel 16kHz capture, hardware AEC & DoA direction |
| Expressive Eye Displays | Dual Waveshare 1.28" GC9A01 LCD (SPI) | 240x240 circular panels for natural pupil & gaze animation |
| Mobility Chassis | 3-Wheel Holonomic (Ø142mm Base) | L-type 520 geared motors with clamped couplers |
| Safety & Navigation | RPLIDAR C1 + VL53L0X ToF + ICM-42688-P | 2D planar scanning, floor edge detection, pick-up sensing |
| Spatial Intelligence | Physarum Adaptive Network Engine (Rust) | Dynamic conductivity field mapping & candidate corridor discovery |
| Cloud AI Engine | Google Gemini Live (Multimodal WebSocket) | Continuous bidirectional voice stream & visual context reasoning |
From verified module validation to scalable cloud deployment.
Integrated dual Sony cameras, reSpeaker DoA acoustic gaze, SPI dual-eye animation, and initial Gemini Live streaming prototype.
Closing the loop on the 3-wheel omni base, integrating real-time IMU pickup detection, and implementing independent hardware fall-prevention stop.
Executing large-scale cloud spatial sweeps, model latency benchmarking, and deploying pilot units to domestic households for trial feedback.
Custom integrated mainboard PCB layout, durable injection molded consumer enclosure, and enterprise fleet management on Google Cloud.
Interested in early trial units, co-developing interactive companion experiences, or technical partnership? Contact our engineering team.