ActirionRobo

An AI-driven 3D robotic action learning and control platform with real-time visualization.

Star Repository • Problem • Solution • Architecture • Capabilities • Benchmarks • Installation

---

Problem Description

Robotics development often suffers from fragmented systems where learning, control, and visualization are handled in isolation. This creates challenges in:

Understanding how AI decisions translate into robot actions

Evaluating learned behaviors beyond raw accuracy

Scaling experimental robotics systems toward industry workflows

---

Solution Overview

ActirionRobo is a modular AI-centric robotics platform that unifies:

Action learning from data

Task-level scheduling and execution

Real-time 3D visualization

System analytics and benchmarking

---

Core Capabilities

AI-Based Action Learning

• Implemented behavior cloning models using PyTorch to learn robotic joint trajectories
• Transformed human–object interaction data into robot joint-level actions
• Enabled smooth and continuous motion prediction

Task Scheduling & Control

• Queue-based scheduler supporting pick, place, move, sort
• Autoplay, step-wise execution, pause, and reset functionality

Dual Control Modes

• Manual joint-level control via UI
• Autonomous AI-driven execution with safe interpolation

Real-Time 3D Visualization

• Live robotic arm animation
• Trajectory and state rendering synchronized with execution

Analytics & Monitoring

• Inference latency tracking
• Task execution timelines
• System resource monitoring

---

System Architecture

//"""""""""" Architecture Updated Shortly """"""""""""

End-to-End Workflow

sequenceDiagram
    participant H as Human / Sensors
    participant FE as Feature Engineering
    participant ML as AI Action Learning Model
    participant JP as Joint-Level Prediction
    participant TS as Task Scheduler
    participant ENV as Robot Environment
    participant VIS as 3D Visualization & Analytics

    H-->>FE: Interaction + State Data
    FE-->>ML: Normalized Features
    ML-->>JP: Learned Actions
    JP-->>TS: Control Commands
    TS-->>ENV: Execute Tasks
    ENV-->>VIS: State + Metrics
    VIS-->>H: Visual Feedback

Loading

---

Benchmarks & Experiments

Evaluation Metrics

• Trajectory Error (MSE)
• Task Completion Rate
• Execution Smoothness
• Inference Latency
• Joint Stability

Experimental Setup

• Fixed task sequences across runs
• Identical environment initialization
• Controlled random seeds for reproducibility

Sample Results

Experiment	Task Type	Avg MSE	Completion Rate
Exp-01	Pick & Place	0.021	96%
Exp-02	Sorting	0.034	91%

---

Planned Extensions

• Reinforcement Learning (Gym / Gymnasium)
• ROS 2 integration
• Physics-aware collision handling
• REST / WebSocket APIs
• Multi-robot coordination
• Cloud-based experiment execution

---

Tech Stack

• AI & ML: Python, PyTorch, NumPy, Pandas
• Visualization: Streamlit, Plotly
• System Engineering: Modular Python architecture
• Monitoring: psutil, logging utilities

---

Installation & Execution

git clone https://github.com/dhakarshailendra829/AI_Robot_3D_Simulator.git
cd AI_Robot_3D_Simulator
pip install -r requirements.txt
streamlit run app.py

Project Status

ActirionRobo is an open-source, academic, and experimental platform intended for learning, research, and system prototyping.
It is not affiliated with any commercial robotics product.

---

👤 Author

Shailendra Dhakad
AI • Robotics • Software Systems Engineering