Our three takeaways from the #AOMT2023 Autonomous Machinery Conference
Algoryx’s Urban Wikman and Nelummali M. participated in the Autonomous Off-Highway Machinery Technology Hybrid Summit which was recently held in Berlin, Germany.
This conference gathered leaders and engineers from heavy machine manufacturers, academic experts, and technology suppliers.
We asked Urban and Nelummali to take all their impressions from #AOMT2023 and condense them into three key takeaways.
1. ) Autonomy is a trending topic, but Automation is the main focus for many heavy machine manufacturers
❝The word ‘autonomous’ was prominent in the title of the conference and many of the sessions. But many of the companies we met and listened to were effectively working on automation initiatives, rather than actual autonomy.
The distinction between autonomy and automation is that an automated system is built to perform a process without human assistance, while an autonomous system is designed to handle uncertainties in the environment and compensate for failures without external aid.
To some extent, it is reasonable to focus on automation capabilities first, as a step on the ladder to full autonomy. But in many aspects, automation and autonomy are two distinctly different things that apply to different scenarios and comprise different technologies.
The companies who build heavy machines, and the companies who use them, will be wise to think long and hard about automation vs. autonomy in their innovation and operations.❞
2. ) Training algorithms on hardware from scratch is still common practice
❝Many heavy machine manufacturers are investing in serious, long-term programs to develop autonomous control systems for their machines and vehicles.
These manufacturers are well aware that electrification and machine autonomy are the two big technology shifts that they must keep up with.
But we find it notable that many R&D departments are training the machine learning algorithms on the actual physical machines, rather than in simulation.
We believe that the development of autonomous control systems must start with simulation. This is the only way to generate the huge amounts of data needed to train the neural network. Simulation is also the only way to train the autonomous system in environments and scenarios that are too dangerous to set up in real life.
Recommended by LinkedIn
But for many machine manufacturers, it is challenging to adapt the workflow and tools that are needed to work with simulations. Control system engineers are close to the hardware machines but distanced from the digital models.
Bridging this gap is obviously an important task for us at Algoryx .❞
3. ) Regulation and legislation is a big story for the coming years
❝One of the highlights of the conference agenda was the presentation by Naïo Technologies ’ Cédric Seguineau about the upcoming Machinery Product Regulation from the European Union, which is expected to apply from mid-2026.
These new rules are an essential step to ensure that autonomous machines will be safe and secure and to establish legal clarity.
Adapting to the new regulations will surely pose challenges for machine manufacturers. Especially since the details of the regulations are not fully defined yet.
In the next three years, compliance will be a big issue for machine manufacturers. It will be interesting to see if regulatory pressures will deepen interest in AI safety throughout the industry.
It will also be interesting to watch out for innovations in the testing and validation of autonomous machines.❞
Summary of the Autonomous Off-Highway Machinery Technology Hybrid Summit (February 23-24, 2023)
Did you participate in the #AOMT2023 conference? Please feel free to share your own reflections in a comment! 💬