Exact mathematical navigation — zero training, zero weights
| Test | Scenarios | Obstacles | Success |
|---|---|---|---|
| Base | 10 | 4 fixed | 100% |
| Random 100 | 100 | 0–12 | 100% |
| Mega 1000 | 1,000 | 0–15 | 100% |
| High-Density 1000 | 1,000 | 8–15 | 100% |
| TOTAL | 2,110 | — | 100% |
No machine learning. No statistical approximation. Pure mathematics — same input = same output, every time.
Exact trigonometry: computes the direction vector from robot to target. Speed scales with remaining distance — decelerates naturally on approach.
Inverse-square repulsion — the same physics law that governs electromagnetism and gravity. Radial evasion + tangential contouring around obstacles.
Fixed heading/avoidance blend. Precision approach mode activates near target with obstacles to contour around final barriers.
Exacts wheel velocity conversion via differential drive kinematics. v_left = forward — angular × wheel_base/2.
Every decision is cryptographically sealed. Each seal includes the previous seal — creating an unbreakable audit chain. 100% repeatability: same scenario produces identical SHA-256 on every run.
6 levels tested: CLEAN to ULTRA (40cm position deviation, 11.5° heading deviation). 360 runs, 100% success. Immune to real-world sensor entropy.
OpenCL on AMD Iceland + Intel HD 620. Obstacle force computation parallelized across all obstacles simultaneously. Crossover at 1,000+ obstacles.
Full ROS 2 node wrapper. Subscribes to /odom (pose) + /scan (LIDAR). Publishes to /cmd_vel (wheel velocities). Deployable to any differential drive robot.
Complete URDF model, arena world file, and ROS 2 launch file. Drop-in simulation environment for testing before hardware deployment.
Every scenario produces identical SHA-256 on every run. Two independent runs of the same scenario generate the exact same cryptographic seal. This is mathematically impossible for any ML-based system — and it is the definitive proof that the ZH-81 is pure computation, not statistical approximation.