Where this conversation describes AI being used to build or deliver a product. Part of the AI in practice tracker.
Engineering & simulation
Machine learning was used to develop control models for soft robots by mapping inputs to deformations, allowing the system to predict how the robot body would behave when actuated without relying on complex, inaccurate physics-based mathematical models.
“So that's why machine learning was kind of good for these kind of robots. So you are actually able to develop models faster. So models certainly means like you need to have an idea about what will happen to your body once I actuate it.”
Engineering & simulation
A neural network was trained using camera data to learn the mapping between input signals and physical deformations, which was then used to develop controllers for gripping objects with soft robotic fingers.
“So you look at use cameras to see what how it actually deforms. Collect a lot of data. So that's the key in nation learning. You collect a lot of data to show this is how it should deform, and then tell a new network to kind of find the mapping between this input and this deformations.”
Manufacturingpick speed doubled from 6 months ago
Ocado Technology implemented a robotic picking system using machine learning and computer vision to achieve successful grips on diverse grocery items without prior knowledge of the object, enabling faster picking speeds.
“One way in which Akado technology has approached this problem. is through the development of an advanced robotic picking system, it uses a suction cup. This system utilizing that cutting age machine learning in computer vision technologies that enable to achieve as a successful grip.”
Engineering & simulationdeep reinforcement learning
Acquired company Kindred Systems utilized deep reinforcement learning and machine vision systems to develop state-of-the-art robotic manipulation technology for autonomous systems operating at scale.
“So, kindred has state of the art robotic manipulation technology with advanced capabilities of deep reinforcement learning, machine vision systems, gripper technologies, celebration, supervised autonomy, to operate autonomous systems at scale.”