Robotec.ai, Liquid AI, and AMD display agentic AI in a dynamic warehouse. Supply: Robotec.ai
Robotec.ai has demonstrated a totally autonomous warehouse robotic. The corporate, which makes a speciality of simulation methods for testing and deploying robotics, is collaborating with Liquid AI and Superior Micro Units Inc.
The robotic makes use of agentic AI to dynamically plan and execute duties in actual time with out reliance on hard-coded scripts, mentioned Robotec.ai.
The robotic is powered completely by AMD Ryzen AI processors and Liquid AI’s next-generation LFM2 imaginative and prescient language fashions (VLMs). These basis fashions mix notion, reasoning, and pure language understanding to interpret instructions, detect security hazards comparable to spills or blocked exits, and autonomously execute corrective actions.
In depth testing in simulation platforms has improved system efficiency, Robotec.ai mentioned. Simulation helps validation of embedded AI on actual {hardware}, avoiding the prices and dangers of bodily testing. It’s a step towards a way forward for reasoning robots that may intelligently reply to the altering setting round them, the firm claimed.
AMD, Robotec.ai, and Liquid AI showcased agentic AI in robotics at ROSCon 2025 in Singapore, the place attendees might see stay demonstrations at AMD’s show in Sales space 17/18.
Liquid AI mannequin integrates notion, reasoning, and language
Bodily intelligence has the potential to create immense worth by way of environment friendly platforms like cell manipulators, mixed with agentic AI that integrates basis fashions and dependable robotics. Robotec.ai mentioned that its new demonstration consists of an autonomous cell robotic (AMR), powered by agentic AI that’s operating on AMD silicon, working inside a warehouse with blended site visitors.
It completes duties specified by people utilizing pure language and adapts to altering circumstances by way of replanning. The versatile robotic additionally serves as an inspection agent, alerting operators at any time when sudden occurrences or questions of safety are detected within the warehouse space.
On the core of this method is Liquid AI’s LFM2-VL, a VLM designed for embedded, real-time intelligence, Robotec.ai defined. It integrates notion, reasoning, and language understanding right into a single multimodal basis mannequin optimized for AMD {hardware}.
To tailor the mannequin for agentic robotics downstream duties, Liquid makes use of simulation-derived artificial information supplied by Robotec. The corporate mentioned this allows domain-specific fine-tuning and robustness in complicated industrial environments.
LFM2-VL interprets visible scenes, performs context-aware reasoning, and plans goal-driven actions solely on-device. Its effectivity and responsiveness allow the robotic to function safely and autonomously, Liquid AI mentioned.
Robotec.ai added that AMD’s processor proved to be a superb “mind” for agentic AI whereas additionally able to operating the robotics software program stack in parallel. “It’s quick, compact, and environment friendly, with unbelievable efficiency in each velocity and energy effectivity, as measured by clear metrics comparable to tokens, latency, and throughput,” the corporate mentioned.
Initially demonstrated on an AMD Ryzen processor, Robotec.ai plans to transition to an AMD embedded x86 resolution within the close to future.
Robotec.ai works with AMD on HiL simulation
Robotec.ai has collaborated with AMD hardware-in-the-loop (HiL) simulation, boosting testing and efficiency. The corporate has created an HiL simulation operating a ROS 2 stack and AI inference on a single Ryzen system.
On this configuration, the simulation runs on a separate AMD-powered laptop and delivers a digital setting that Robotec.ai asserted is indistinguishable from the bodily world to the robotic.
The HiL interface then connects the robotic’s sensor and actuator indicators on to the simulator’s inputs and outputs, permitting the identical management logic to be utilized beneath reasonable, reproducible circumstances.
It helps the transition of the HiL setup to the real-world setup, enabling fast OEM prototyping and demonstrating the answer’s robustness. In the long term, HiL can speed up the analysis and improvement section of innovation, mentioned Robotec.ai.
