TRISTONES AI ROBOTICS LABS LLC filed two foundational patent applications with the United States Patent and Trademark Office (USPTO) in February 2024, covering the core technical innovations that distinguish the company's approach to AI-powered service robotics.
Patent 1: LLM-to-Motion Bridging System
The first application covers TRISTONES's proprietary method for translating high-level natural language action plans generated by large language models into safe, smooth, and mechanically feasible actuator commands in real time.
This is harder than it sounds. LLMs produce outputs in natural language or structured text — "pick up the glass from the table and hand it to the user gently." Converting that instruction into the precise sequence of joint angle commands, force limits, approach vectors, and collision-avoidance behaviors that a physical robotic arm must execute requires a translation layer that is both fast (sub-100ms latency for interactive applications) and robust (safe across the enormous variety of real-world physical configurations).
TRISTONES's bridging architecture achieves this through a combination of learned motion primitives, real-time physics constraint solving, and a novel uncertainty quantification mechanism that degrades gracefully when the LLM's instruction is ambiguous rather than guessing dangerously.
Patent 2: Multi-Modal Emotion Recognition for HRI
The second application covers a system for recognizing human emotional states from multi-modal inputs — combining facial expression analysis, vocal tone and prosody, body language and movement patterns, and contextual situational inference — and using those recognized states to modulate robot behavior in real time.
This capability is central to what makes TRISTONES robots feel different from conventional service robots. A robot that can recognize that a user is frustrated, anxious, or distressed — and adjust its communication style, task prioritization, and physical proximity accordingly — delivers a qualitatively different experience. In care environments, this distinction is not merely cosmetic; it has measurable effects on user wellbeing and care outcomes.
IP Strategy
The two February 2024 filings join a growing portfolio of granted patents covering the company's 6-DOF compliant arm design, Semantic SLAM navigation module, and Predictive Fall Prevention algorithm. TRISTONES's IP strategy is focused on protecting the fundamental innovations in the AI-physical interface — the company's core technical moat — rather than pursuing broad defensive filings.
"Our patents protect the things that actually make our robots work better," said the company's IP lead. "Not just the shapes of parts."