TECHNOLOGY
ARTERA Intelligence Engine
(*ARTERA: AI Real-Time Engine for Reasoning Augmentation)
CORNERS' Real-time Safety Intelligence Enhancement AI Engine, achieving the core requirements for on-site safety ADX
Core requirements for on-site safety ADX
■ Golden-time
→ High-level real-time capability, ultra-low latency
■ Liability
→ High reliability, high precision (zero false positives/misses)
■ Complex strike
→ Integration of Sensing-Judgment-Execution
ARTERA Core Technology
ARTERA Detect
Achieving over 90% accuracy in real-time incident detection through the integration of vision, congestion level, sound (gunshot), and robot sensor data
- Simultaneous processing of over 60 channels per GPU
- NMS-Free high-density stability
- Edge AI lightweight optimization
- Utilization of existing CCTV infrastructure
- Frequency-based shockwave pattern recognition
- Self-learning of indoor reverberation data
- Real-time tracking of multiple sensor locations
- Operation in closed network environments
- Immediate detection of victim locations
- Spatial adaptation and dynamic control in extreme on-site environments, and manager-AI collaboration
- Sensing data noise control
ARTERA Determine
Generating optimal evacuation routes within approximately one second by recognizing complex situational context through Digital Twin and ontology-based inference
- Generating optimal paths considering crowd flow control and bottlenecks
- Path generation within approximately 1 second, with continuous reflection of on-site conditions
- First response
ARTERA Direct
Supporting on-site operations through selective monitoring and immediate situational dissemination of field conditions
ARTERA CoCo
Through multi-agent orchestration, Humans and AI collaborate to autonomously operate unmanned facilities and automatically record and manage incident response procedures
ARTERA ADX Platform
ADX platform developed to apply the ARTERA AI Engine, which consists of Detect, Determine, Direct, and CoCo, to real-world field sites
4-Layer Structure of the ARTERA Agentic AI Platform
Detect · Determine · Direct
+ CoCo
• Perception → Judgment → Action cycle is the agent's autonomous loop
• Implemented via OODA and Cyber-Physical Loop
Ontology-based Digital Twin
(Builder·Watch·Simulator)
• Semantic foundation structuring the physical world into formal data
• Connects AI judgment to facts (same role as Palantir Ontology)
InteleTwin · InteleGuide
· InteleAgent
• Domain-specific productization of core technologies on a common platform
• Each product is a collection of mission-performing agents
Smart Agent Devices · Quadruped Robots
· Vision Sensors
• Real-time Physical AI guidance in the safety ADX field and 'Detection-Judgment-Generation-Dissemination' situational change guidance
• Support for industrial standard protocols
➊ Technology Layer: 3D + CoCo (3 Core Tech.)
(Brain: ARTERA Engine)Classification system for safety-based core technologies accumulated by Corners (Detailed functions/models exist under each classification)
CoCo AI Collaboration
Multi-agent orchestration (Planner → Executor → Critic) · LLM-independent deterministic FSA(Finite State Automata) execution engine
Orchestration between agents on a common platform
Core technologies integrated into each product
➋ Application Layer: 3 Main Products (Set of agents performing missions)
InteleTwin
Digital Twin Integrated Monitoring
(Video Object Detection · Neuro-symbolic Situational Judgment · Edge Computing)
InteleGuide
Real-time Evacuation Route Guidance
(Hazard Recognition · Route Generation · Dynamic Guidance Functions)
InteleAgent
LLM/VLM Specialized
(Coordinator, Designer, Executor, Monitor Agents)
Digital Twin/Ontology Call ↔ Data/Status Information Provision
➌ Service Delivery Platform Layer
Semantic foundation commonly applied to most projects
Synchronization · Control (Cyber-Physical Loop)
➍ Physical Layer
Physical targets that Digital Twin synchronizes and controls in real-time
External Tool Integration
CoCo · Product layer calling/orchestrating via tool-calling(MCP) (Bidirectional)