Future Electronics and Microchip — Northern Lights Development System

Future Electronics and Microchip have developed a multi-board demonstration system for an Artificial Intelligence (AI) application for robotics, motor control, and image recognition.

The new Northern Lights demonstration system includes image-processing and neural network elements. A Video Board uses an image sensor to view a single digit, and performs AI at the edge to recognize the digit and send an instruction to the Motor Control Board. This board then directs a robotic arm to move an object to a location marked with the same digit. A Display Module shows the image from the camera, and displays a signal to show when the system has recognized a digit.

The Northern Lights system features Microchip products which perform the key roles, including implementing the neural network inference engine, image processing and display control, motor control and driving, and power regulation.

At the heart of the Video Board is a low-power, secure Microchip IGLOO2 FPGA, the M2GL025-VF256, which features 28,000 logic elements, extensible to 150,000. The FPGA, which implements a Convolutional Neural Network (CNN) model for image recognition, features a RISC-V CPU implemented in software, a scaler for the display output, and a UART interface.

The Video Board also includes an ON Semiconductor MT9M114 CMOS image sensor, a Microchip SAM E51 microcontroller based on an Arm^® Cortex^®-M4 core, and a Microchip MIC33M350 power module.

The Display Module houses a 160px x 128px OLED display, the WiseChip UG-6028GDEBF02. The module also has a standard video connector which is reversible for viewing flexibility.

The Motor Control Board is based on a Microchip dsPIC33EP128GM604 Digital Signal Controller (DSC). It communicates with the Video Board via a Microchip MCP2561 CAN transceiver. This board is also powered by an MIC33M350 module. The dsPIC33 device supplies a PWM output signal to control the movement of the robotic arm.