AI-ML & ROBOTICS

Introduction

We offer an AI-DL Lab designed to explore and develop skills in Machine Learning, Deep Learning, Deployment/Inference on Embedded Hardware, and Robotics. The lab provides configurable options tailored for both research activities and student training.

The AI–DL Laboratory is equipped with advanced hardware and software setups to facilitate skill development in Machine Learning, Deep Learning, Deployment Techniques on Embedded GPU Boards, and Robotics.

• A structured set of experiments designed to cover both fundamental and advanced AI concepts.
• Access to Embedded GPU boards for hands-on practice in deployment techniques

AI/DL LAB HARDWARE SETUP:

AI/DL Machine Setup: AI Server

The AI/DL machine is equipped with high-performance GPU cards, along with the required tools, libraries, frameworks, and sample datasets, offering a complete platform for exploring critical AI skills such as Machine Learning, Deep Learning, Deployment/Inference, and Robotics. This setup provides an end-to-end environment for both training and application development.

Technical Specifications:

Processor: 2x 5th Gen Intel Xeon Scalable Silver 4516Y, 24 Core, 48 Thread, 2.2GHz Base, 3.7GHz Turbo, 45MB Cache, 185W TDP
Motherboard: 2x CPU sockets, 32 DIMM slots
RAM: 8x 64GB DDR5 RDIMM ECC (Total 512GB DDR5)
Storage (OS): 1x 960GB SSD NVMe
Storage (Data): 2x 8TB 7200 RPM SATA Enterprise HDD
LAN: 2x 10G Gigabit Ethernet
Ports: Front 1×PCIe x8, Rear 1×PCIe x16, w/8 × 3.5" bays (8 SATA / 2 NVME)
Management Port: 1×MGMT (back)
Form Factor: 4U Rack Type Server Chassis (supports up to 8 GPU Cards)
Display & Input: HDMI compatible, USB Mouse & Keyboard

GPU Configurations

Option	GPU Model	Memory per GPU	Configuration	Total GPU Memory
1	NVIDIA RTX Pro 4000 Blackwell	24 GB	4× / 8×	96 GB / 192 GB
2	NVIDIA RTX Pro 4500 Blackwell	32 GB	4× / 8×	128 GB / 256 GB
3	NVIDIA RTX 5000 ADA	32 GB	4× / 8×	128 GB / 256 GB
4	NVIDIA RTX Pro 5000 Blackwell	48 GB	4× / 8×	192 GB / 384 GB
5	NVIDIA RTX 6000 ADA	48 GB	4× / 8×	192 GB / 384 GB
6	NVIDIA RTX Pro 6000 Blackwell	96 GB	4× / 8×	384 GB / 768 GB

Embedded Inference Hardware: Embedded GPU

The setup includes a dedicated embedded GPU hardware platform that enables learners to understand and practice the process of deploying and inferring trained models/networks on embedded systems.

This hardware is bundled with a customized suite of tools and libraries, conveniently provided on an external hard drive/SSD, ensuring a smooth and efficient learning experience.

Research Areas

Machine Vision
Robotics
Deep Learning Model Inference
Machine Learning
Medical Imaging
Gaming
Virtual Reality
NLP & Many More...

Embedded GPU with 6 core NVDIA

Technical Specifications:

6-core NVIDIA ARMv8.2 CPU
384-core Volta GPU @ 1100MHz with 64 Tensor Cores
Dual Deep Learning Accelerator (DLA) engines
7-way VLIW Vision Accelerator
8GB LPDDR4x
MIPI CSI-2 lanes
UART, SPI, I2C, I2S, CAN, GPIOs

Embedded GPU with 12 core NVDIA

Technical Specifications:

12-core NVIDIA ARMv8.2 CPU
2048-core Ampere GPU @ 1.3 GHz with 64 Tensor Cores
Dual Deep Learning Accelerator (DLA) engines
Programmable Vision Accelerator
64GB LPDDR5
64GB eMMC
UART, SPI, I2C, I2S, CAN, GPIOs

Camera Setup

This hardware setup includes a range of camera systems with supporting accessories, designed to facilitate the implementation of AI skills in the field of computer vision and image processing. The setup is bundled with multiple types of cameras, as outlined below:
• Thermal Camera – For infrared imaging and heat signature analysis
• 3D-Stereo Camera – For depth estimation and 3D vision applications
• Night Vision Camera – For low-light and dark-environment imaging
• IP Camera (Wireless) – For network-based vision applications
• USB Camera – For standard imaging and prototyping

Robotic Setup

This hardware setup includes a Robotic Arm and Robotic Car, which can be utilized to implement AI skills through the Embedded GPU Kit and the Robotic Operating System (ROS). The platform enables learners to design, build, and test AI-driven robotic applications.

Offering hands-on experience in areas such as autonomous navigation, robotic control, and intelligent automation.

AI / ML Based Jetson Nano Robotic Arm

This hardware setup includes Robotic Arm which can used to implement AI skills using Embedded GPU Kit & Robotic Operating System (ROS) by building applications.
Applications:
• Simultaneous movement of dual robotic arms.
• Gesture recognition, • Color interaction, • visual positioning, • garbage sorting, • catch game, • face tracking, • blocks stack and others AI vision game play

AI / ML Based Jetson Nano AI Robotic Car

Applications: • Al racing car powered by N-VIDIA Jetson Nano.

DonKeyCar Project:

DoneKeyCar utilizes deep learning neural network framework Keras/TensorFlow, together with computer vision library OpenCV, to achieve self driving.

SLAM Lidar Mapping:

Mapping with odometer, IMU, lidar, EKF, etc.

AI/DL LAB SOFTWARE SETUP:

The setup is pre-configured with Ubuntu OS (16.04 / 18.04) and comes with a comprehensive collection of libraries, utilities, tools, SDKs, and datasets, ensuring a ready-to-use platform for AI and Robotics development.
Essential Utilities
• CUDA, cuDNN, TensorRT
Machine Learning Libraries
• Vowpal Wabbit, XGBoost
• NumPy, Scikit-learn, Pandas, and other essential Python libraries
Deep Learning Frameworks
• NVIDIA DIGITS
• TensorFlow
• Caffe, Caffe2
• PyTorch, Torch
• Theano
Pre-loaded Datasets
• ImageNet
• CIFAR-10
• KITTI
(ready for out-of-the-box development and experimentation)

Robotics Tools & Frameworks

• ROS (Robotic Operating System)
• OpenAI (Reinforcement Learning & Q-Learning)
• Simulation and visualization tools: Gazebo, RViz, MoveIt!, Autoware.ai, Apoll

Pitsco TETRIX PRIME Robot for FPGA Based Reconfigurable I/O(RIO)device

The TETRIX Prime Robot, when combined with NI FPGA Based RIO device provides a powerful and flexible learning platform for exploring robotics, mechatronics, and control systems. Built on a durable and modular construction system, the TETRIX Prime kit allows students to quickly design and assemble a variety of robotic models. With NI RIO's real-time processing, I/O capabilities, and LabVIEW integration, students can program and control their robots with industry-standard tools

ROVER VEHICLE ASSEMBLY

Students are tasked with building a robot that can drive, turn and process commands from a computer via Wi-Fi. Students will later add more advanced functionality such autonomous operation. This first project is a versatile and fun starting point for a variety of mechatronics design projects.

BALANCING ARM ASSEMBLY

The TETRIX Prime Balancing Arm is an assembly that demonstrates how control concepts taught in engineering classes can be applied. Students will learn how to integrate sensors, servos and a PID control algorithm with a robot that balances a ball in a position specified by the user.

SELF BALANCING ROBOT

The self-balancing robot is a complex closed-loop control system that autonomously balances itself in place. Students build the mechanical structure, create a system that read multiple sensor inputs, and implement a PD control algorithm to keep the robot upright.

Nyro Ned2-6-axis collaborative robotic arm

Nyro Ned2 is a 6-axis collaborative robotic arm designed for education, research, and prototyping. It is compact, affordable, and safe for desktop use, making it ideal for teaching robotics, AI, automation, and Industry 4.0 concepts.
Key Features:
•A 6-axis collaborative robotic arm designed for education, research, automation, and industry 4.0 prototyping.
• Built around open-source tech: runs Ubuntu 18.04 with ROS Noetic/Melodic, and supports Blockly, Python, C++, ROS, MATLAB, and more .
• Specs: 300 g payload, ~0.5 mm repeatability, ~49 cm reach, Raspberry Pi 4 controller, Wi-Fi/Ethernet/USB ports Applications:
• Programming with Blockly – Simple sequences (e.g., wave hand, draw a shape).
• Coordinate Systems – Understanding joint, Cartesian, and tool coordinates.
• Basic Pick & Place – Moving an object from point A to B using a gripper.
• Real-time Tracking – Following a moving object with vision guidance.

AI basedIndustry 4.0 Robot

Robot System with Assembly and Sorting

This system is fully functional models of actual applications, mimicking hybrid, real life, and industrial automation scenarios. A wide variety of project assignments and learning objectives help students to build hybrid systems by integrating important automation technologies, such as
• Robotics
• Vision system
• Electrical
•Mechanical
• Sensors
Applications:
This robot performs vision-based quality checks and intelligently sorts work pieces by material type into storage or rejection bins.

SCARA (Selective Compliance Assembly Robot Arm)

A SCARA (Selective Compliance Assembly Robot Arm) is a high-speed industrial robot widely used for pick-and-place, assembly, and material handling tasks. Known for its rigid vertical movement and flexible horizontal reach, the SCARA robot delivers precision, reliability, and efficiency in automation processes. Its compact design and ease of integration make it an ideal choice for manufacturing, electronics assembly, packaging, and quality inspection applications.

Technical Specifications:
• Payload: At least 4 kg
• Reach: At least 600 mm
• Z-Stroke: 200 mm Minimum

Applications:
• Pick-and-Place: Fast and precise object transfer
• Assembly Tasks: Screwing, inserting components, or dispensing adhesive
• Packaging & Palletizing: Placing products into boxes or trays
• Machine Tending: Loading/unloading parts from CNC or testing equipment
• Quality Inspection: Moving sensors/cameras to specific positions for part inspection

Pick and Place Delta Robot

The Delta Robot is a high-speed, precision robotic system widely used for fast pick-and-place operations, particularly in packaging, sorting, and assembly tasks. Its lightweight parallel-arm structure allows for extremely rapid movements with high accuracy, making it ideal for handling small, lightweight components.

Technical Specifications:
• Dimensions (L x W x H): 1000 x 800 x 1700 mm
• Aluminum profile: Tabletop profile – 40 x 160 mm; Supporting profile – 40 x 40 mm
• Grid spacing (slot to slot): 40 mm
• Profile groove width: 8.3 mm
• Profile plate connectors: Length 55 mm, thickness 5 mm

Applications:
• High-Speed Pick-and-Place: Sorting items from a conveyor using vision
• Food & Beverage Industry: Picking candies, bakery items, fruits, or packaged goods
• Pharmaceutical Automation: Vial handling, blister packing, and rapid sorting

Industrial IoT(IIoT)-FPGA Based

Harness the power of real-time intelligence, reliability, and flexibility with ourFPGA-based IIoT solutions. Designed for modern industries, our offering enableshigh-speed data acquisition, seamless connectivity, and smarter decision-making through a complete hardware-software ecosystem.

Our Solution Package Includes:

• FPGA-Based Reconfigurable I/O Controller & Chassis: High-performance, flexible control system with modular I/O for diverse industrial needs.
• LabVIEW Software with Real-Time & FPGA Module: Powerful programming and monitoring environment for data processing, analysis, and rapid system deployment.
• ThingsSpot IIoT Toolkit: Ready-to-use IIoT integration toolkit for secure cloud connectivity, analytics, and remote monitoring.
• Essential Sensor Set: Reliable basic sensors to enable accurate data acquisition and smart industrial applications.

• IIoT Software Module: The ThingsSpot IIoT Toolkit is a software add-on for LabVIEW that provides an Industrial Internet of Things (IIoT/IoT) solution to help you control and gather data from industrial controllers, PLCs, sensors, and IoT devices.
• The add-on supports cross-integration with NI embedded controllers including CompactRIO Controllers and CompactRIO Single-Board Controllers.
• You can log data to local databases and view data visualized on configurable dashboards that offer real-time control and monitoring through a web GUI.
• With the ThingsSpot IIoT Toolkit, you can also push the data to cloud services such as IBM Cloud, Microsoft Azure, Amazon Web Services (AWS), and more.