An efficiency revolution with computer vision and AI
Over the course of my career working in robotics and automation, I have witnessed the incredible advancements of AI and computer vision firsthand. At Ocado Technology, our approach to these technologies has been to learn fast, constantly experimenting, learning from mistakes, and ultimately pushing the boundaries of what’s possible for our use case. Today, I’d like to share some insights into how some of the most advanced technologies we are building are reshaping the online grocery operation, the challenges we’ve encountered along the way, and what the future holds.
Enhancing operations: computer vision for automation technology
At the core of our automated customer fulfillment centres, hundreds of retrieval bots navigate 3D storage grids, fetching containers filled with the items that customers have ordered. This highly efficient system is orchestrated by an AI-powered “air traffic control” that communicates with each bot in real-time, ensuring seamless operations. However, as with any physical system, occasional mechanical issues arise. This is where one of our applications of computer vision comes into play.
We’ve built advanced AI and computer vision systems which monitor the footage from 360° cameras that cover the whole grid. These cameras are used to detect and locate any bots that require maintenance attention. AI allows us to create exclusion zones around any bots experiencing an issue, minimising disruptions and maintaining the flow of operations.
Another significant advancements we’ve made in the computer vision space is the development of On-Grid Robotic Pick (OGRP) systems. These robotic arms are installed directly on the grid, picking and packing customer orders with precision. The variety of products—up to 50,000 different items—poses a unique challenge, but by using cutting-edge AI and computer vision algorithms, we’ve developed a system that can handle the vast diversity in weight, size, shape, and packaging. Our approach involves deep neural networks, imitation learning, and reinforcement learning, creating a robust and ever-improving system.
Overcoming challenges in integration
Transitioning from a controlled lab environment to real-world operations is one of the most intriguing challenges we’ve faced. In the lab, we simulate realistic conditions, but the sheer scale of production reveals scenarios that are nearly impossible to replicate during testing. Events that occur once in a million instances are rare in the lab but can be frequent enough in production to warrant attention.
To address these challenges, we’ve developed capabilities that augment our systems’ autonomy with human teleoperation. This not only ensures continuous operations but also contributes to the AI’s learning process, enhancing its ability to handle complex tasks over time. Each robot arm benefits from the collective experiences of every other arm in operation, leading to exponential improvements in performance.
Not just efficiency; the customer wins too
Automation is not only about efficiency; it’s also about delivering an exceptional customer experience. Computer vision-powered automation means that products can be picked, packed, and delivered with minimal human error. Our AI-driven inspection models operate independently from our control models, verifying that the correct items and quantities are selected in real-time.
Our adaptive packing algorithms also ensure that products are handled with care, prioritising the safety of fragile items while optimising speed for more durable goods. This balance is continuously refined to meet our retail partners’ specific needs, ensuring that customers receive their orders in optimal condition.
What’s next in the computer vision space?
Looking ahead, we see two major trends shaping the future of automation: broader applicability and lower costs. Machine learning is enabling us to create more generalisable and adaptive systems, capable of handling increasingly complex scenarios. Our OGRP system, for example, can manage tens of thousands of different products, a testament to the power of ML in expanding the capabilities of automation.
Cost reduction is another significant trend. By leveraging ML, we can use off-the-shelf hardware, making advanced automation more accessible and affordable. We’re also exploring the potential of foundational models, which have already transformed natural language processing and image recognition. These models could revolutionise robot control, opening new frontiers in automation and further enhancing our ability to deliver top-tier solutions.
In conclusion, the integration of computer vision with automation technologies is not just a technological advancement; it’s a revolution that is starting to redefine the future of distribution. At Ocado Technology, we’re excited to continue leading this charge, developing innovative solutions that not only enhance operational efficiency but also elevate the customer experience to new heights. The future is exciting, and we’re just getting started.
Dive deeper with more AI insights:
If you're interested in AI adoption you can find more insights on our AI Campaign Week page, here:
Stay up to date with techUK AI Adoption work:
To help promote greater levels of responsible AI adoption, techUK is working alongside its members and key stakeholders across the AI ecosystem to demonstrate the significant benefits of this technology for both the economy and society. The best way to stay up to date with our work is to visit our AI Adoption Hub.
Call for AI Adoption case studies:
As part of techUK’s 2024 AI Campaign Week, we are launching a call for techUK member case studies, to demonstrate how organisations are tackling the barriers to AI adoption to maximise AI's potential. These case studies will be showcased to UK organisations that are also embarking on their AI adoption journeys, providing them with practical examples and insights to guide their Putting AI into Action efforts. Learn more here.
techUK - Putting AI into Action
techUK’s Putting AI into Action campaign serves as a one stop shop for showcasing the opportunities and benefits of AI adoption across sectors and markets.
During this campaign, techUK will run a regular drumbeat of activity, including events, reports, and insights, to demonstrate some of the most significant opportunities for AI adoption in 2024, as well as working with key stakeholders to identify and address current barriers to adoption.
Visit our AI Adoption Hub to learn more, or find our latest activity below.
Upcoming AI Adoption events
AI risks for energy networks – workshop
techUK, 10 St Bride Street, EC4A 4AD RoundtableLatest news and insights
Get our tech and innovation insights straight to your inbox
Sign-up to get the latest updates and opportunities from our Technology and Innovation and AI programmes.
Contact the team
Usman Ikhlaq
Usman Ikhlaq
Usman joined techUK in January 2024 as Programme Manager for Artificial Intelligence.
His role is to help techUK members of all sizes and across all sectors to adopt AI at scale. This includes identifying the barriers to adoption, considering solutions and how best to maximise AI's potential.
Prior to joining techUK, Usman worked as a policy, government affairs and public affairs professional in the advertising sector. He has also worked in sales and marketing and FinTech.
Usman is a graduate of the London School of Economics and Political Science (MSc), BPP Law School (GDL and LLB) and Queen Mary University of London (BA).
When he isn’t working, Usman enjoys spending time with his family and friends. He also has a keen interest in running, reading and travelling.
Learn more about our AI Adoption campaign:
Authors
Alex Harvey
With over 14 years of experience in leading advanced robotics, machine learning, and simulation teams, Alex played an important role in developing Ocado’s proprietary automation warehouse technology and cutting-edge IP. Most recently this includes leading and integrating all of Ocado’s technology acquisitions and investments, and the development of it’s state-of-the-art robotic picking and packing system.