JetRacer ROS AI Kit A - a kit for building an autonomous robot with NVIDIA Jetson Nano

Description

A kit for building an autonomous racing robot based on the NVIDIA Jetson Nano B01 development kit. It consists of a racing chassis, an 8MP camera with Sony IMX219 sensor, a 2.4/5 GHz Dual-Band WiFi wireless communication module, IMU sensors, a laser scanner and necessary accessories.

The chassis is a design based on the Ackerman principle with a 333-rpm 37-520 drive motor and a servo-driven front torsion axle. The chassis has a heavy-duty metal gearbox and a double-locking differential. The chassis skeleton is made of high-quality carbon fiber and sits on four shock absorbers to stabilize it when driving over demanding terrain. Additional components of the drivetrain include incremental encoders with Hall sensors, which provide the feedback necessary for the algorithms used in odometry.

The kit offers a wealth of sensors. The 12-meter-range RPLidar laser scanner provides 360-degree scanning for a contour map around the robot, and the built-in 9-axis position sensor, which uses an extended Kalman filter to combine data with odometry sensors, allows the robot's position to be determined precisely.

Responsible for image processing is the MX219 sensor camera with a resolution of 3280x2464 pixels, which is ideally suited for the development of vision algorithms with the OpenCV suite such as object recognition, target tracking and autonomous driving.

The robot also offers speech recognition and voice control capabilities, as well as sound generation. It uses two high-quality MEMS microphones and a speaker for stereo playback. It supports speech synthesis, detection and recognition algorithms for human-robot interaction. The software has the ability to detect the audio stream and remove the muted part, so that the robot picks up only the part that the user is speaking. In addition, it offers both text-to-speech and speech-to-text processing. It can work with Google Assistant in English and IFLYTEK in Chinese.

The robot is controlled by an NVIDIA development kit with an ARM Cortex A57 CPU running at 1.43 GHz. It is equipped with a Maxwell graphics chip consisting of 128 cores supporting CUDA technology. The chip has 4 GB of DDR4 RAM, Ethernet and HDMI interfaces and two MIPI-CSI connectors. The package is designed especially for graphics design and processing of very large amounts of data. It allows you to get acquainted with the technique of programming GPUs.

The Jetson Nano module is the smallest device of the Jetson platform. It provides performance to run modern AI workloads in a small, low-power format. It provides 472 GFLOPs of computing power for high-speed operation of modern AI algorithms. It can run multiple neural networks in parallel and simultaneously process data from multiple high-resolution sensors, making it perfect for applications such as basic network video recorders (NVRs), home robots and smart gateways with full analytical capabilities.

The robot's expansion board features a motor controller, OLED display, charging and battery monitoring circuit, and a cage for 3 18650 battery cells. The user will also find an audio amplifier, USB audio card, IMU sensors and MEMS microphones. The module is structed by a dual-core Raspberry Pi RP2040 microcontroller, which features better real-time performance and higher control accuracy, and is responsible for data collection and motion control.

The robot is supported by a wealth of software, including ROS, an open-source operating system that includes a collection of software libraries and robot design tools. With DonKeyCar, it uses the Keras/TensorFlow framework for deep learning along with the OpenCV computer vision library to achieve the highest possible degree of driving autonomy, while a rich set of sensors combined with the Kalman filter (EKF) gives it the ability to implement SLAM algorithms.

The manufacturer provides users with numerous examples and extensive documentation to easily get started with the kit. More information and materials can be found on the product page.

Features

• NVIDIA Jetson Nano B01 controller with 4GB RAM
• RP2040 microcontroller
• Speaker
• 2 x MEMS microphone
• IMU sensor
• RPLidar laser scanner
• Support for SLAM algorithms
• IMX219 8MP camera, 160° FOV
• OLED 0.91" 128×32 px
• 2.4G/5G dual-band WiFi, Bluetooth 4.2 AC8265 module
• Ackerman chassis
• Power supply: 12.6V, 3 x 18650 battery (not included)
• 37-520 motor with metal gearbox and encoder
• 9kg/cm servo
• Aluminum chassis
• Front bumper
• Support for ROS, DonkeyCar, JetRacer libraries

Kit contains

• Jetson Nano Dev Kit
• 64GB microSD card
• MicroSD card reader
• JetRacer ROS case
• Acrylic piece
• Camera mount
• JetRacer ROS expansion board
• RPLIDAR A1
• IMX219-160 camera
• Wireless-AC8265
• USB adapter
• 2 x Motor with encoder
• Motor cover
• Servo mount
• MG996R servo
• 4 x 65mm tire
• Cooling fan
• 2 x EVA felt pad
• Wireless gamepad
• 2 x PH2.0 6PIN cable
• 2 x Antenna
• 2 x IPEX to SMA adapter
• 12.6V 2A battery charger
• Pack of tools and accessories
• Chess board
• Set of screws