CSIRO, Features, Innovation, Manufacturing Opinion

CSIRO to assist automated system innovators with an enhanced development platform

ROS is a major open-source software initiative that emerged out of Stanford University in the USA during the mid-2000s.

During more than two decades with CSIRO I’ve witnessed and been part of numerous digital innovations across Australian manufacturing and other industries. In the majority of cases, I’d class these as being part of what is known as the third industrial revolution – computerisation.

Author: Simon Dunstall – principal research scientist, CSIRO

We are now in the midst of the fourth industrial revolution – Industry 4.0.

Robotics and automation are central to Industry 4.0 and are major CSIRO R&D priorities because we’re both voracious users and cutting-edge developers of them. Globally and domestically the manufacturing industry is at a challenging point. Imperatives for companies, industries and countries include achieving net zero emissions, improving sustainability, and contributing to a circular economy.

Robotics, autonomy and Artificial Intelligence (AI) are key for ongoing business competitiveness, and for establishing more sustainable manufacturing systems. Pertinent examples of robotics, autonomy and AI in the manufacturing and logistics context are autonomous ground or aerial vehicles, intelligent assistive cobotics, and AI-assisted machine and process controls.

These can underpin step changes in competitive advantage. They are also complex innovations requiring dependable hardware and software components. I’m a strong advocate for using Robotics Operating System or ROS as the software platform upon which to base many of these kinds of innovations.

What is ROS?

Some in the manufacturing industry will be familiar with ROS and its current version ROS2. ROS is a major open-source software initiative that emerged out of Stanford University in the USA during the mid-2000s.

Additional to operating robotics and automation, ROS is a useful toolkit for complex sensing, computer vision, data fusion and real-time AI. ROS is applied in scientific, industrial, and aerospace domains. ROS or ROS2 is not actually an operating system in the standard sense of computer operating systems such as Windows or Linux. ROS is a software framework and suite of libraries used for controlling and coordinating robotics system elements.

The software executes within a mainstream computer operating system on any device capable of running it, including cameras, robot arms and machine controllers, within a centralised, distributed and/or edge computing manner. ROS2 is now the preeminent open-source option for programming robotics, automation and sensing. It is predicted that 55 per cent of commercial robots shipped in 2024 will have at least one Robot Operating System package. (Source: ABI Research, 2019).

At CSIRO we have used ROS and ROS2 within autonomous vehicles, robots constructed from flexible materials, real-time distributed video streaming, and real time object recognition systems. We have also used it in integrated laboratory robotics and sensing systems, additive manufacturing robotics control and real-time sensing, and multi- camera high precision measurement systems.

ROS2 can handle greater volumes of larger and/or more complex data in closer to real time compared to the MQTT-based systems that are popular for Industrial IoT (noting also that ROS2 can interoperate with MQTT, e.g., for ROS2 nodes to exchange MQTT messages with IoT sensors).

ROS2 offers a non-proprietary and expanding development platform, efficient and effective solution creation, and long- term support from a large pool of international practitioners.

Taking on ROS2

In a manufacturing context, CSIRO is using ROS2 to integrate robot arms, cameras and scanners, additive manufacturing and machining tools, and computer vision and other AI. ROS2 enables us to manage chemical and materials laboratory automation, process sensing and control, and the pre-commercial development of new robotic additive manufacturing work centres. We are also pursuing many other ROS2 applications within CSIRO in agriculture and other domains besides the manufacturing domain.

In our experience, and from what others have written, we know that understanding and using ROS2 can be challenging. This is especially the case for smaller teams using the platform for the first time. The learning curve can be very steep. Within CSIRO there are many ROS2 success stories but also many burned fingers from insufficient assistance in the initial stages of adopting ROS2.

Having recognised a number of challenges, the CSIRO has implemented the ROSella program to help industry.
Having recognised a number of challenges, the CSIRO has implemented the ROSella program to help industry.

Kick-starting a journey by contracting or hiring ROS2 developers is a great option. Accessible and professional sources of expertise and help around ROS2 are indeed available. However, few companies in Australia have been long-term adopters of ROS2, and experienced ROS2 engineers are now in great demand globally. The creates a workforce undersupply issue, and I see this as having three critical dimensions.

Firstly, a lack of people with practical ROS2 knowledge leads to a continued reliance on proprietary, legacy, and out-of-date approaches to automation, sensing and robotics. This constrains Australia’s manufacturing innovation success in terms of technical performance, interoperability, and product longevity.

Secondly, it locks the available robotics software engineering labour force into a low productivity regime. The old ways require more intensive effort in developing and (especially) maintaining solutions. Moreover, they put out of reach the opportunity to freely leverage the vast global capacity which is directed at open platforms including ROS2.

Thirdly, the Australian robotics community is probably too small to properly support unduly divergent and proprietary approaches to robotics. Australia needs workforce mobility between projects, products, and employers. Similarity and familiarity of solutions that are developed across the manufacturing sector and the Australian innovation system is also required.

We need to be able to concentrate our education and upskilling effort.

It expected that some elements of these dimensions will be addressed or dissipate in future. For now, for a practical ROS2 journey, this puts the emphasis back onto developer communities and self- education approaches.

The global online developer community is a great asset for any development team, but not sufficient to ensure initial or ongoing success. Vibrant geographic and sector-specific communities are crucial. Resources such as do-it-yourself ROS2 project kits, designed to deliver early wins, are also required.

To the best of my knowledge, a community for ROS2 in Australia mainly exists among R&D institutions.

A broad and national ROS2 community is only nascent in Australian advanced manufacturing. I know of many advanced manufacturing robotics and machine developers, Industrial IoT product innovators, and system integrators in Australia who are aware of ROS2. The majority however are presently deferring ROS2 adoption for want of stronger innovation support and better entry points that assure success.

The ROSella initiative

CSIRO is seeking to play its part in turning this around via an initiative that we have named ROSella. Initially incubated inside CSIRO as a step- change in how we approach robotics and automation development, we are launching ROSella nationally in the second half of 2023. ROSella is intended as an open national community around ROS2 with a strong advanced manufacturing applications focus.

Curating and maintaining a subset of ROS2 and other software modules that interoperate and address major use cases in manufacturing is a key plank of ROSella. In our experience, ROS or ROS2 can become difficult to get up and running when third party software (proprietary or open) is needing to be integrated with it.

Getting Windows deployments of ROS2 to work has also been troublesome on occasion. Commercial developers in particular need confidence that teething troubles and core platform interoperability is assured, hence ROSella’s focus is on a core ROS2 subset.

Easing the adoption barrier is helped by the availability of hardware and software starter kits that get ROS2 working out-of-the-box. These are demonstrators, confidence builders, and seeds for more complex systems.

At CSIRO we have developed a low-cost computation box and camera kit that allows a ROS2 demonstrator to be built and operated by “newbies” in a single afternoon. Several such units are being made available across our organisation for hire by R&D teams beginning ROS2 journeys. Extending this service to others in the innovation system is part of ROSella’s brief.

To be part of a ROS2 system, each intelligent device in the system needs to implement a ROS2 software interface, which is only sometimes natively implemented out-of-the-box. Furthermore, the international ROS2 community can lag in developing ROS2 software interfaces for newly released devices, most notably for cameras of various kinds.

In-silico device representations are also necessary for simulating systems, for example in ROS’ own Gazebo simulator prior to building them and this extends the interface development effort. CSIRO is equipping the ROSella community with tools and resources to speed up the device interface and in-silico emulator development process.

ROSella is also a channel for advanced IP developments out of CSIRO and elsewhere to reach mainstream robotics, automation and sensing applications. Key present examples are in new methods for multi camera system calibration and control, and in enabling real-time video streaming and object-detection AI through coupling ROS2 with the popular GStreamer libraries.

Nucleating an energised and distinctly Australian innovation community around ROS2 is another major aim of ROSella.

We have learned from experience and observation that community building is not a simple task. Facilitating the open exchange of ideas and value, at self- sustaining scale, takes careful curation and significant resourcing.

For ROSella this will come in part through the activities like the ROS2 starter kits. It will also need incumbent ROS2 devotees, including CSIRO, to come together and co-invest the time, money and energy that is needed to support Australia’s ROS2 adoption.

If this collaborative approach speaks to your business, then contact Simon Dunstall or Christian Ruberg on 1300 363 400, or email fdmf@csiro.au.

Send this to a friend