πŸ€– Physical AI Β· On-device VLM

λΉ„λ“œλž˜ν”„νŠΈ Γ— Boston Dynamics Spot

μƒμš© 4μ‘± 보행 λ‘œλ΄‡μ— λΉ„λ“œλž˜ν”„νŠΈ μ˜¨λ””λ°”μ΄μŠ€ VLM을 μ–Ήμ–΄, μŒμ„±μœΌλ‘œ μ œμ–΄ν•˜κ³  λŒ€ν™”ν•˜λ‹€

🧠 μ˜¨λ””λ°”μ΄μŠ€ VLM νƒ‘μž¬
πŸŽ™οΈ μŒμ„± λͺ…λ ΉΒ·μ œμ–΄
πŸ’¬ μŒμ„± λŒ€ν™”
πŸ”Œ μ¨λ“œνŒŒν‹° μ†Œν˜• λͺ¨λ“ˆ λΆ€μ°©
πŸ“ 둜컬 μΆ”λ‘ 

λΉ„λ“œλž˜ν”„νŠΈλŠ” λ³΄μŠ€ν„΄ λ‹€μ΄λ‚΄λ―ΉμŠ€(Boston Dynamics)의 4μ‘± 보행 λ‘œλ΄‡ β€˜Spot’에, μ¨λ“œνŒŒν‹°λ‘œ μ†Œν˜• μ˜¨λ””λ°”μ΄μŠ€ λͺ¨λ“ˆ(λΉ„λ“œλž˜ν”„νŠΈ VLM νƒ‘μž¬)을 별도 λΆ€μ°©ν–ˆμŠ΅λ‹ˆλ‹€. 이λ₯Ό 톡해 λ‘œλ΄‡μ„ μŒμ„± λͺ…λ ΉμœΌλ‘œ μ œμ–΄ν•˜κ³ , λ‘œλ΄‡κ³Ό μŒμ„±μœΌλ‘œ λŒ€ν™”ν•  수 있게 ν–ˆμŠ΅λ‹ˆλ‹€.

λ‘œλ΄‡ μ œμ‘°μ‚¬μ˜ κ°œμ‘°λ‚˜ ν΄λΌμš°λ“œ 의쑴 없이 β€” μ™Έμž₯ μ˜¨λ””λ°”μ΄μŠ€ AI만으둜 μƒμš© λ‘œλ΄‡μ— μ‹œκ°Β·μ–Έμ–΄Β·μŒμ„± 지λŠ₯을 λΆ€μ—¬ν•œ 피지컬 AI(Physical AI) μ„±κ³Όμž…λ‹ˆλ‹€. λΆ€μ°©ν˜• κ΅¬μ‘°μ΄λ―€λ‘œ λ‹€λ₯Έ λ‘œλ΄‡ ν”Œλž«νΌμœΌλ‘œμ˜ ν™•μž₯도 κ°€λŠ₯ν•©λ‹ˆλ‹€.

VIDRAFT attached a compact third-party on-device module β€” carrying a VIDRAFT vision-language model (VLM) β€” to Boston Dynamics’ Spot quadruped robot, enabling voice command & control and voice conversation with the robot.

No manufacturer modification and no cloud dependency β€” an external on-device AI alone brings visual, language and voice intelligence to a commercial robot. Because it is an attach-on module, it can extend to other robot platforms.

Powered by VIDRAFT Β· Darwin VLM Β· On-device Physical AI  |  vidraft.net
β€» β€˜Spot’ and β€˜Boston Dynamics’ are trademarks of their respective owners; used here descriptively to indicate the host robot platform.