Refl3x develops bio‑inspired robotic systems that separate mechanical structure, peripheral sensing, and control logic into distinct layers. Their architecture converts raw sensor data into high‑level contact states—such as slip, overload, or jam—providing robots with a reliable, real‑time understanding of physical interactions for more predictable and safe operation.
Funding
Funding not disclosed
Founders
Product
Problem
Robotic systems often lack reliable, real‑time awareness of physical interactions, leading to unsafe commands that can cause slip, overload, jams, or collisions. Existing control architectures rely on high‑level software to make perfect decisions, which is unrealistic in uncertain or degraded sensing conditions.
Solution
REFL3X introduces a biologically inspired, layered architecture that separates mechanical design, peripheral sensing, and a deterministic reflex kernel. The mechanical “body” defines force limits, contact points, and failure modes, while the “nerves” layer converts raw sensor data into simple contact states such as slip, overload, jam, or stable contact. These states feed a reflex layer that intercepts every movement command, instantly limiting, correcting, or rejecting actions that would violate safety constraints. By enforcing physical safety at the boundary between high‑level intelligence and actuators, the system provides a reliable, real‑time picture of machine‑environment interaction, enabling stable operation even when perception or planning modules make errors.
Target Audience
Primary customers are developers of autonomous robots and manipulators—including industrial arms, dexterous hands, collaborative robots, drones, mobile robots, and autonomous vehicles—that require guaranteed safety and stability during physical interaction.
Features
- Three‑layer architecture (Body, Nerves, Reflex) mirroring biological systems for predictable physical behavior
- Peripheral sensors that abstract raw data into discrete contact states (slip, overload, jam, stable)
- Deterministic reflex kernel that enforces force, torque, and motion limits before commands reach actuators
- Automatic command modification or rejection (limit force/velocity, hold position, back off, or abort) in unsafe situations
- Compatibility with any AI, planning, or robotics stack; can be implemented in software, firmware, or embedded hardware
- Handles degraded sensing by defaulting to safe motion limits or hold‑still behaviors