Skip to content

Welcome to Signalbotics Documentation

Complete guide to training vision-language-action models and reinforcement learning algorithms for robotics. Follow our holistic workflow from data collection to deployment.

Complete Robot Learning Workflow

graph LR
    A[Data Collection] --> B[Dataset Prep]
    B --> C[Simulation]
    C --> D[Training]
    D --> E[Evaluation]
    E --> F[Deployment]

    style A fill:#e1f5ff
    style B fill:#e1f5ff
    style C fill:#fff4e1
    style D fill:#ffe1e1
    style E fill:#e1ffe1
    style F fill:#f0e1ff

This documentation provides a step-by-step guide through the entire process:

  1. Data Collection - Collect demonstrations or prepare datasets
  2. Dataset Preparation - Format data using LeRobot standard
  3. Simulation Setup - Configure IsaacSim, IsaacLab, or Newton
  4. Training - Train VLA, RL, or IL models
  5. Evaluation - Test and validate your policies
  6. Deployment - Deploy to real robots

What You'll Learn

Training Methods

Vision-Language-Action (VLA) Multi-modal models for language-conditioned robot control. Perfect for tasks requiring natural language specification and zero-shot generalization.

Reinforcement Learning (RL) Optimize policies through environment interaction. Includes PPO, SAC, TD3, and more. Ideal when you can specify reward functions.

Imitation Learning (IL) Learn from expert demonstrations. Best for tasks with available demonstrations but hard-to-specify rewards.

Essential Tools

LeRobot Dataset Format Standardized format for robotics datasets with multi-modal support, efficient storage, and Hugging Face integration.

Simulators - IsaacSim - Photorealistic rendering for perception tasks - IsaacLab - Fast GPU-accelerated RL training - Newton - New physics engine for maximum speed

Quick Start

Choose your path based on your needs:

Path 1: Learn from Demonstrations (1-2 weeks)

Collect demos → LeRobot format → Train BC policy → Deploy
Start with Imitation Learning →

Path 2: Reinforcement Learning (2-4 weeks)

Setup simulator → Define reward → Train PPO/SAC → Deploy
Start with RL →

Path 3: Language-Conditioned (3-6 weeks)

Collect multi-modal data → Train VLA → Test generalization → Deploy
Start with VLA →

Robot Learning (Main Tab) - Complete workflow from data to deployment - All training methods in one place - LeRobot dataset format - Simulator guides (IsaacSim, IsaacLab, Newton)

Products - Future product offerings

Services - Consulting and training services

About - Company info and contact

Getting Started

New to robot learning?

  1. Read the Getting Started Guide
  2. Review Best Practices
  3. Follow the complete Robot Learning Workflow

Ready to dive in?

Jump directly to any step in the workflow: - 1. Data Collection - 2. Dataset Prep (LeRobot) - 3. Simulation Setup - 4. Training (VLA/RL/IL) - 5. Evaluation - 6. Deployment

Key Features

✓ Holistic Documentation - Complete workflow in one place ✓ Multiple Training Methods - VLA, RL, and IL approaches ✓ Standardized Format - LeRobot dataset specification ✓ Simulator Guides - IsaacSim, IsaacLab, Newton ✓ Code Examples - Practical implementations throughout ✓ Best Practices - Industry-tested recommendations

Support

Start your robot learning journey today! Follow the structured workflow and achieve results faster.