Skip to content Skip to footer

AI in Robotics: Advancements and Challenges in the Industry

Artificial intelligence (AI) has significantly impacted the robotics industry, enabling robots to act autonomously and respond to their external environment in real-time. This has opened up new possibilities for robotic automation, particularly in manufacturing and logistics sectors where uncertainty and variability are common features. However, the integration of AI in robotics also raises questions around safety, certification, and regulation.

AI algorithms have enabled robots to act autonomously, raising questions around safety. However, robot applications comprise many different levels of software, and currently, the safety layer is hard-coded, such as ‘stop if an object is less than 10cm away’. The global robotics industry is concerned about the current drafts of the European Machinery Product Regulation and the AI, particularly with regard to the proposed requirement for mandatory third-party certification of AI-enabled robots. This would impact any company selling robots on the European market, disadvantaging European companies, especially SMEs and start-ups.

AI has opened up new possibilities for robotic automation, enabling robots to sense and respond autonomously to their external environment in real-time. For example, AI enables robots to identify objects to be picked from an unsorted bin or automatically identify welding points on a new part. AI also holds potential for reducing the time and resources needed to programme and re-task robots. Programming and integration account for 50-70% of the cost of a robot application, making automation economically unviable for many small-to-medium-sized companies or for larger manufacturers and wholesalers with high product variance.

However, AI is not necessarily a prerequisite to enable robots to respond in real-time to their environment. Many ‘pick-and-place’ applications, in which the robot identifies an object to be picked and determines how to approach and grasp the object, do not require AI. The greater the level of variability and uncertainty, the more likely it is that AI algorithms will bring cost-benefits over traditional, deterministic programming.

Recent work has improved robots’ ability to reason, enabling them to use chain-of-thought prompting, a way to dissect multi-step problems. The introduction of vision models, like PaLM-E, has helped robots make better sense of their surroundings. And RT-1 has shown that Transformers, known for their ability to generalize information across systems, can help different types of robots learn from each other.

Google DeepMind has introduced RT-2, a vision-language-action (VLA) model that can directly output robotic actions. RT-2 is trained on text and images from the web, enabling it to transfer knowledge from web data to inform robot behavior. This new approach has shown promise for robots to more rapidly adapt to novel situations and transfer learned concepts to new situations.

AI has significantly impacted the robotics industry, enabling robots to act autonomously and respond to their external environment in real-time. However, the integration of AI in robotics also raises questions around safety, certification, and regulation. The global robotics industry is concerned about the current drafts of the European Machinery Product Regulation and the AI, particularly with regard to the proposed requirement for mandatory third-party certification of AI-enabled robots. Despite these challenges, AI holds potential for reducing the time and resources needed to programme and re-task robots, making automation economically viable for many small-to-medium-sized companies.

Leave a comment

0.0/5