This week, at the Embedded Vision Summit, we are pleased to announce that we are joining the Arm AI Partner Programme. We are already running on several Arm-based processors, including Raspberry Pi 4 and NVIDIA TX2, both of which we are using to demonstrate our Spatial AI SDK on our virtual booth at the Summit. We are also part of Qualcomm’s ecosystem for its RB5 Robotics Platform. Joining the Arm AI Partner Programme will help open doors to work with even more Arm-based processors. As a key player behind the explosion in on-device AI, Arm plays a crucial role in the development of capable, commercially viable robots as solutions to many of today’s challenges and this partnership will significantly accelerate the development and adoption of SLAMcore’s algorithms.
Effective and accurate Spatial AI demands that algorithms and hardware are closely optimized. The computational overhead of delivering the complex calculations needed for robots to successfully position themselves, map and understand the physical world can be huge. To work effectively in the size, power and cost-constrained world of mobile autonomous devices SLAM algorithms must be exactly tailored to the embedded processors on which they run.
This presents a challenge to those developing SLAM for the industry. Robot designers need a different combination of sensors, processors, batteries and hundreds of other components to meet their product requirements. Creating an algorithm that works with just one processor would limit you to a very narrow niche. But trying to optimize for a wider range of hardware can lead to generic solutions that are neither accurate nor robust enough for real-world applications.
One solution is to offer a consultancy-based approach in which the software providers undertake to optimize their algorithms for specific hardware configurations – for a fee. But this is time consuming, difficult to scale and can be cost prohibitive for many smaller robot developers. From the outset we’ve taken a different approach. By identifying, partnering and working closely with a small but growing number of the leading hardware firms SLAMcore offers optimized solutions for the best and most widely used processors and sensors. By choosing these combinations, designers can be confident that both hardware and software are perfectly matched to deliver highly accurate SLAM in an efficient and cost-effective way.
If you are interested in finding out more about the hardware we support, or if you are a hardware vendor interested in joining our partners programme please come and visit us at the Embedded Vision Summit (which is virtual and taking place until 25th September), or get in touch here.