MIRAI is an AI-powered computer vision platform that enables industrial robots to detect objects, estimate pose, and adjust trajectories in real time without CAD models or manual calibration. It handles reflective, transparent, and dirty parts, integrates via OPC‑UA and ROS APIs, and allows on‑site model retraining to support high‑mix, low‑volume production.
Funding
$30M raised to dateRaised to date based on public sources. This may differ from the amount the company actually raised and is based only on what is publicly available on the internet.
Founders
Product
Problem
Manufacturers often face production variance—such as reflective, transparent, or dirty parts—that defeats conventional robot vision systems, which typically require CAD models, precise camera calibration, and static lighting conditions. This limits automation of tasks like insertion, testing, and rack hanging, leading to lower throughput and higher labor costs.
Solution
Micropsi Industries’ MIRAI platform delivers AI-powered computer vision that enables industrial robots to perceive and adapt to real‑time changes in part appearance and environment. Leveraging deep‑learning convolutional neural networks, MIRAI performs on‑board object detection, pose estimation, and trajectory adjustment without any CAD data or pre‑defined visual features. The system tolerates reflective surfaces, transparency, dust, and moving objects, allowing robots to maintain sub‑millimeter accuracy across variable production cycles. Deployment is streamlined through a plug‑and‑play software stack that integrates with standard robot controllers via OPC‑UA and ROS‑compatible APIs, and engineers can retrain models on‑site in minutes to accommodate new parts or process updates. This reduces integration time, improves cycle‑time consistency, and expands the range of feasible automation applications.
Target Audience
Primary customers are manufacturing plants and system integrators that automate high‑mix, low‑volume production lines—particularly in electronics assembly, consumer appliance testing, and logistics rack handling—where part variability challenges traditional vision solutions.
Features
- End‑to‑end deep‑learning vision pipeline with real‑time inference (< 30 ms latency) for object detection, pose estimation, and dynamic path planning.
- Calibration‑free operation: automatic camera parameter estimation and adaptive illumination handling eliminate the need for manual setup.
- Robust perception under challenging conditions, including high reflectivity, transparency, dust, grime, and moving parts.
- Continuous learning workflow: on‑site data capture and rapid model retraining via a web‑based UI, supporting zero‑downtime updates.
- Integration layer offering OPC‑UA, ROS, and proprietary robot controller APIs for seamless command execution and feedback loops.
- Scalable cloud‑optional analytics module that aggregates inspection data for quality monitoring and predictive maintenance.
- Built‑in safety compliance with ISO 10218 and IEC 61508, providing fail‑safe vision checks and emergency stop triggers.
- Comprehensive SDK with Python and C++ bindings for custom application development and third‑party system integration.