Session: 07-03-03: Design and Control of Robots, Mechanisms and Structures

Paper Number: 95438

95438 - Flexible Low-Level Control Software Framework for Achieving Critical Real-Time Deadlines 

In this work, a low-level framework is proposed to simplify software development for Hardware Abstract Layered (HAL) control systems, verify profile task completion, and identify networking methods for accurate real-time communication between devices. Many robotic systems use multi-layered control architectures which have a centralized processor responsible for high-level decisions and several low-level controllers responsible for local tasks. These low-level tasks include sensor acquisition, networking, enforcing safety limits, and executing robust input action, which are performed at different speeds depending on the robotic system. Missing a single deadline for one of these tasks on an actuated system could quickly lead to unwanted and unstable action with fatal consequences. Therefore, it is imperative to design robotic systems with a guarantee that tasks are completed on time.

Furthermore, several software and hardware constraints complicate the design of low-level controllers for robotic systems. Low-level controllers are often implemented using single-threaded microcontrollers, only capable of executing a single task at one time. Microcontrollers are inherently opaque, and their internal states cannot be easily observed during runtime. Traditional networking strategies between microcontrollers and high-level processors are inefficient due to unnecessary overhead procedures. HAL programming is entirely different depending on the microcontroller making it tedious and time-consuming to transition between platforms. Multi-layer controlled robotic systems require a low-level framework with timing guarantees for network communication and task execution to ensure good performance and safe action. The developed framework achieves the required deadlines within these constraints by implementing a real-time operating system to manage task priorities and measure processor utilization. The programming framework is a general approach which is irrespective of the hardware, giving more flexibility to a designer during a development cycle.

We implement this framework on a distributed microcontroller system composed of Texas Instruments TM4C123GXL TIVAs for a humanoid robot. The robot’s high-level controller decides whole-body dynamic motion, with low-level controllers responsible for each individual joint. A real-time operating system (TI-RTOS) ensures crucial deadlines are met and provides advanced task profiling on the low-level controllers. The TI-RTOS diagnostic tools allow the designer to optimize the low-level system for efficient task completion. All microcontroller software is unified into one program and uses initiation files from the high-level processor to configure each individual TIVA depending on its location on the robot. The EtherCAT communication protocol is used to avoid delays from IP address routing. Overall, our proposed framework overcomes the major challenges of writing low-level control software so that development is less time-consuming, simpler to manage, and easier to validate. Further, this work is generalizable for many kinds of robotic systems and applications that use microcontrollers within a multi-layered control architecture.

Keywords: Software Framework, Networking, Microcontroller, Control System, Communication, Operating System

Presenting Author: Connor Herron Virginia Polytechnic Institute and State University (Virginia Tech)

Presenting Author Biography: Connor Herron received two bachelor’s degrees from Virginia Tech in Mechanical Engineering and Electrical Engineering in 2020. Connor joined the Terrestrial Robotics Engineering and Controls (TREC) laboratory at Virginia Tech in Summer 2017 as an undergraduate researcher. His undergraduate research focused on a range of topics surrounding robotics from PCB design, sensor integration, analog and digital filtering, system identification, and robust controller design for multi-layered systems. He has been a part of the new full-sized humanoid robot made entirely of 3D printed components and in Fall 2020 began his Graduate research studying higher-level controls applications in robotics. His research interests involve control engineering, humanoid robotics, exoskeletons and assistive devices, virtual reality technology, and haptic feedback.

Authors:

Nicholas Tremaroli Virginia Polytechnic Institute and State University (Virginia Tech)
Maxwell Stelmack Virginia Polytechnic Institute and State University (Virginia Tech)
Connor Herron Virginia Polytechnic Institute and State University (Virginia Tech)
Bhaben Kalita Virginia Polytechnic Institute and State University (Virginia Tech)
Alexander Leonessa Virginia Polytechnic Institute and State University (Virginia Tech)