AI Module Library

Robora offers a curated suite of pretuned and edge-optimized Ai Modules: These AI Modules cover a broad spectrum of robotics and IoT needs.

Each model will be versioned on-chain via the Model Registry, with metadata on performance benchmarks, resource needs, and compatibility (edge vs. cloud). Users can select or upgrade models in the dApp, ensuring the right “brain” for each mission.

Offered AI Module Library

To cover a broad spectrum of robotics and IoT needs, Robora will offer a curated suite of pre‑tuned and edge‑optimized AI models:

• Vision & Perception

  • YOLOv8-Tiny for real‑time object detection on edge devices.

  • DeepLabv3+ for semantic segmentation tasks in inspection and mapping.

  • Meta SAM (Segment Anything Model) for interactive region‑of‑interest selection and annotation.

• Localization & SLAM

  • ORB-SLAM3 for robust visual‑inertial SLAM on mobile robots.

  • RTAB-Map for RGB-D mapping and loop closure in indoor/outdoor environments.

  • Google Cartographer for high-fidelity 2D/3D mapping.

• Control & Path Planning

  • RRT (Rapidly-Exploring Random Trees) for efficient motion planning.

  • PPO (Proximal Policy Optimization) reinforcement-learning models for adaptive control.

  • MPC (Model Predictive Control) modules for trajectory optimization.

• Language & Instruction

  • GPT-4[Vision] and Gemini-lite for natural-language command parsing and reasoning.

  • LLaMA-2 (fine-tuned) for on-device language understanding and dialogue.

• Edge-Optimized & TinyML

  • MobileNetV3 and EfficientNet-lite for low-power image classification.

  • TinyML models for keyword spotting and anomaly detection in IoT sensors.

• Domain-Specific Agents

  • CropScout-V1/V2 for precision agriculture tasks.

  • InspectBot-V1 for industrial inspection workflows.