Apple's HALO: Fully Autonomous Perceptive Humanoid Motion From Simulation
CyberRobooo · x · 2026-11-06
Apple's AI research team unveiled HALO, a fully autonomous perceptive humanoid motion system trained from simulation, with paper, video, and code released.
The onboard stack has three tiers:
- Orchestrator (0.5 Hz): an off-the-shelf VLM on the backpack compute unit turns spoken requests ("push box away", "follow me", "sit inside") into subtasks, advancing on success and retrying on failure;
- Generator (3 Hz): from RGB input and the current subtask, a frozen Qwen3.5-2B encoder with a DiT decoder (flow matching, 15 Euler steps) produces 5-second whole-body plans of 150 frames × 36 values, committing the first 0.5 s back as clean tokens;
- Tracker (50 Hz): on embedded compute, follows the plan using onboard depth and proprioception to output targets for 29 joints.
Demo skills include crawling under desks, sidestepping through doors, pushing boxes, following people, and sitting on chairs/sofas, with an 84% success rate on the sit task and switchable real-time chunking at inference.
Related event: Apple Unveils HALO, a Fully Autonomous Humanoid Robot(3 posts)→
More from Embodied
- Modded MacBook Pro gets a working touchscreen — used to play Honor of Kings — xiaohu · 2026-11-06
- Roboticist shares Stanford-collab research directions and BARS talk — Fowe · 2026-11-06
- RAX From Harvard/NVIDIA/Berkeley Lifts Robot Success From 26% to 94% With Frozen Weights — jiqizhixin · 2026-11-06
- Open-Weights Robot-Use Harness: Qwen3.8 Scores 31.39, Beating GPT-6 Astra's 28.97 — aigclink · 2026-11-06
- Dev builds Salvo battleship game for dual-screen iPhone Duo with head-to-head and AI play — HankYeomans · 2026-11-06
- First-hand demo of Wayve AI Driver on San Francisco streets impresses rider — Sethwinterroth · 2026-11-06