• collaborative
• control
• FPGA
• motion
• motor
• Robotics
• vision

Industrial automation is experiencing a revolution in the relationship between robots and humans. Industrial robots are being developed for more flexible and collaborative tasks, and are being designed to work safely side by side with human workers. Lower-force robots may need sophisticated motion control and behavioral algorithms, but are expected to be more compact and lower cost.

In order to meet these needs, a hard processor plus FPGA architecture is proposed for the control architecture of a 4DOF articulated robot.
The FPGA provides extreme flexibility in integration of different modules and in reconfiguration.
It is also a good environment for the development of a motor control system.
The hard processor can handle the high level tasks, and eventually run an operative system.
The implementation presented combines multi-axis motion control with 3D robot vision.
The end effector of the robot manipulator reproduces the same movements of the closest point tracked from the camera.
The processor receives the depth frames from the 3D camera, tracks the position of the closest point for each frame, and sends these points to the FPGA, as points of a path.
The FPGA receives the coordinates of the points and computes position and speed commands for each motor. The motion control algorithm in the FPGA includes kinematics and trajectory planning calculations.

The dissertation starts discussing the organization of an industrial automation hierarchy, the main features of collaborative robots and the requirements that these features impose on the control system. Based on these, the advantages coming from an FPGA based control system are discussed, and an overview of the state of the art in FPGA motor
control is provided.

The theoretic background necessary for the motion control of a 4 DOF articulated robot arm is presented. In particular the geometric based solution of the direct and inverse kinematics problem is provided. Singularities are evaluated from the Jacobian matrix, and different approaches for trajectory planning are presented.

Then, the description of the robot hardware and of the control architecture is presented.
Details are provided about the tasks that are assigned to each part of the system, according to the real-time requirements.

The software development for the hard processor and the FPGA is described in detail.
The environment of the hard processor is composed of the Ubilinux operative system and the libraries for the interfaces handling. The program for the hard processor represents the depth image in the terminal, finds the position of the closest point to the camera and sends the position to the FPGA. The frame rate from the camera is 30 fps, and all the program functionalities run at the same rate.
The development flow for the FPGA is performed with the Quartus II Software. First is designed a System on Programmable Chip composed of a Nios II processor plus memory and peripherals. A custom IP is designed for the generation of PWM signals for the control of the motors. The Nios II software implements the inverse kinematics and the trajectory planning algorithms. The code in the FPGA has also the important role of synchronizing the data position from the hard processor, and sending the commands with the right timing for the motion of the robot arm.

This project shows how the requirements for the control of a collaborative robot are met with an architecture composed of FPGA plus hard processor. The demonstration implements successfully the motion control of the robot arm, integrated with the camera vision, that provides the interactive behavior.
Future developments may include: the integration of force sensors, safety IP, industrial
communication protocols, a different trajectory planning algorithm, a motor control method for bigger industrial robot and the use of a different strategy for the image processing.