AI Grand Prix · 2026 Season

Autonomous drone racing,
engineered for the Virtual Qualifier.

Two-track plan: a deterministic VQ1 completion stack (YOLO + PnP + PID) ships first to lock in a passing submission. APEX perception-aware PPO takes over for VQ2 speed. No GPS. No absolute positioning. FPV + telemetry in, Throttle / Roll / Pitch / Yaw out. Per the confirmed 2026-04-19 AIGP spec.

Series
Anduril × DCL · AI Grand Prix 2026
Prize $500,000
VQ1
Virtual Qualifier 1 — Completion
May 2026 · <10 gates
VQ2
Virtual Qualifier 2 — Fastest Time
Jun–Jul 2026 · <20 gates
Physical
In-person qualifier, no audience
Sep 2026 · California
Final
AI Grand Prix Final
Nov 2026 · Ohio
Detector mAP@50
97.9%
Pipeline target
<50ms
Qualifier rounds
02
Prize pool
$500K

§ 01Season Timeline

VQ1
MAY · Completion
VQ2
JUN–JUL · Fastest time
Physical
SEP · California
Final
NOV · Ohio

§ 02Pipeline

Same input surface and same output surface for both tracks. Only the controller changes.

FPV + Telemetry
sim input
YOLO11n
detector
Keypoints
4 corners
PnP
gate pose
PID / PPO
VQ1 · VQ2
T / R / P / Y
sim output

§ 03Read first

Winning Playbook

MASTER

The opinionated master strategy. Effort budget across stages, reliability math, data pipeline moat, sim-to-real bridge, compute envelope, explicit anti-patterns, scorecard.

how to winread first

Winning Strategy

TACTICAL

Tactical expansion of the playbook. VQ1 vs VQ2 stacks, training pipeline, detector choice, retired components, Sim Day 1 checklist, intel.

updated 2026-04-21

APEX Pipeline

TRAINING

Three-phase training — YOLO11n detector → YOLO11n-pose keypoints → perception-aware PPO. Phase 3 observation-swap flag required for VQ2 transfer.

train_apex.pyobservation-mode

§ 04AI & Architecture

VADR-TS-002 deltas

SPEC · NEW

What changed in the 2026-05-08 spec revision (camera intrinsics, UDP vision stream, ODOMETRY removed) and how the code was aligned on 2026-05-12.

5 deltas11 files updated

AI Integration Plan

PLAN

VADR-TS-002 compliance checklist, 3-tier training strategy, submission pipeline, risk matrix.

19 requirements3 gaps

System Architecture

CORE

End-to-end pipeline diagram, component file map, sensor budget, commands surface.

pipelinefile map

Race Pipeline

CORE

Six states, transition rules, latency budget for a single frame-to-command cycle.

6 states<50ms budget

Pipeline Integration

WIRING

How the APEX detector chain + telemetry adapter + controller bolt into the sim client.

integration

Vision & Detection

CORE

Detector choices (YOLO11n, RF-DETR), dataset layout, evaluation harness, confidence thresholds.

dataset_gates_mega

§ 05Training

Trainer App (local)

NEW

Local Windows / FastAPI dashboard. Live GPU telemetry, one-click controls, run history, live log tail. Double-click launch_trainer_app.bat.

aigp_trainer_app.pylocalhost:8080

Training Runbook

GUIDE

Exact training commands + the overnight_autotrainer.py CLI. Unattended nightly runs with precheck, backup, benchmark, auto-promotion, rollback.

overnight_autotrainer.pyrunbook + CLI

Pre-Simulator Training Plan

PLAN

Two-track plan: VQ1 completion stack + VQ2 APEX PPO with observation swap. Parallel sim instances, dataset expansion for distractors.

7 steps

DCL Gameplay Capture

HOWTO

Auto-label training data by screen-capturing DCL The Game while you play. Closest VQ2-realistic imagery we can get pre-sim. ~3-5K frames per 2-4 hr session.

capture_dcl.pydataset_gates_dcl

Hard-Negative Mining

HOWTO

Feed the detector gate-lookalike images (arches, fans, scaffolding) with empty labels to suppress VQ2 false positives. Keyword-based harvest + auto false-positive discovery.

mine_hardneg.pydataset_gates_hardneg_v2

Obstacle Detection

PLAN

Three-tier plan to close the "nothing handles obstacles" gap the AIGP spec exposes. Headless Playwright + race-r3f.html produces auto-labeled nc=2 training data. Free synthetic imagery.

render_synthetic_dataset.pydataset_gates_obstacles

Local GPU Training

HOWTO

One-command APEX run on the RTX 5080 box (~7.5 hr overnight). Outputs, weights, troubleshooting.

train_apex.pyRTX 5080

Model Evaluation

TOOL

Detector benchmark harness — YOLO11n vs RF-DETR vs U-Net on the same course proxy.

benchmark_models.py

Tuning Reference

REF

Gains, thresholds, timeouts. One source of truth for PID + PPO hyperparameters.

race_config.py

§ 06Submission & Ops

Submission Guide

GUIDE

Windows-only sim, anti-cheat internet, T/R/P/Y outputs, nine-check submission list, portal TBD.

submit_check.py

Getting Started

NEW

Clone, install deps, smoke test, first training run. For contributors joining the repo fresh.

10 min

Quickstart

HOWTO

Minimum viable run: detector weights in, frame in, command out. Single-file reproduction.

1 script

Troubleshooting

REF

CUDA OOM, observation-mode confusion, submission-validator failures, anti-cheat handshakes.

known issues

§ 07Hardware

DIY 5-inch Racer Build

BUILD

Full BOM, frame, FCU, ESCs, motors, cameras. Practice-rig recipe for physical-round prep.

BOM

Archer Shell Design

CAD

Printable carrier shell for the Neros Archer platform (supplied at physical qualifier).

STL · DWG

3D Printed Parts

CAD

Mounts, camera brackets, antenna holders — everything printable for the build.

STL index

RF Controller Picker

REF

Transmitter + receiver combos validated for development. Protocol + latency notes.

ExpressLRS · TBS

Build Guide v3

GUIDE

Assembly steps, soldering order, first-flight checklist. Updated for the 2026 practice rig.

step-by-step

Blueprint Drawings

CAD

Dimensioned drawings for frame + camera mount. Source of truth for BOM disputes.

dimensioned

§ 08Reference

Winning Playbook

MASTER

How we win. Effort budget, reliability math, data pipeline moat, sim-to-real, anti-patterns, scorecard.

strategy coreread first

Active Links

INDEX

Deployed dashboard URLs, Cloudflare Pages, Workers, upstream AIGP resources.

aigp-dashboard.pages.dev

APEX Reference

REF

Phase-by-phase details: detector, keypoints, perception-aware PPO. Observation schemas for both modes.

train_apex.py
AIGP DOCS · v2.1 2026-04-21 · Design System F1.2 · Dashboard