Imitation Learning from Teleoperation
Our Imitation Learning Platform redefines the way robotic systems are developed and deployed. It allows you to directly show a robot how to perform a task, by first performing and recording it several times through Teleoperation. Then, we use cutting-edge imitation learning AI algorithms to teach the robot to perform it autonomously in a few hours.
This method enables seamless replication of many complex tasks, enables anyone to teach a robot what to perform and how to perform quickly. More importantly, this process reduces drastically the time needed to develop new motions: when coding a single motion could take a few weeks or even months, our platform enables to do it in a few days. This makes it easier and faster to adapt robotic systems to new tasks, ensuring they can meet evolving demands efficiently.
Advantages
By observing and mimicking human actions, robots or cobots (collaborative robots) can acquire the skills needed to perform complex manipulations with precision and adaptability. This approach offers several significant advantages:
- Rapid Skill Acquisition: Robots can learn complex tasks quickly, allowing for faster deployment and reduced training time.
- Precision and Adaptability: Robots can handle delicate materials like soft fabrics and complex materials with high precision expanding the reach and scope of robotic systems.
- Sim-to-Real Transfer: Our platform avoids many Sim-to-Real transfer problems, since we learn directly from real-world teleoperation recordings.
- Reduced Human Workload: By automating repetitive and complex tasks, our cobots free up human workers to focus on higher-level activities, promoting a safer and more productive collaboration between humans and machines.
- Seamless Integration: Cobots can easily integrate into various environments, from home settings to industrial workplaces, adapting to different tasks and conditions.
Bin Picking
Our platform revolutionizes bin picking by equipping robotic arms with the ability to learn from human demonstrations. By observing and imitating human movements, our robots achieve exceptional precision and efficiency in handling and sorting complex items from bins, whether it’s classic warehouse packages, small parts or even soft materials.
Our platform significantly enhance the scope of classical bin picking which was so far limited to a subset of objects. With our technology, your robotic arm can adapt to various shapes, sizes, and orientations of items, making it ideal for complex bin picking applications in warehouses and production lines.
Simulation
In Reality
Assembling Multiple Objects Together
Harnessing advanced imitation learning algorithms, our robots excel at assembling multiple components with remarkable precision. By mimicking the detailed actions of human assemblers, our robots can handle complex assembly tasks, ensuring each piece is correctly placed and secured.
Our platform enables automating tasks that were not previously possible and still required human workforce to perform motions, often at the cost of inefficiency and in worst case scenarios, source of dire musculoskeletal disorders.
Kitting
Our platform’s advanced imitation learning capabilities enable robots to efficiently handle kitting processes, where items have to be grouped or packaged together.
When only a few types of objects and packages was previously available to automated kitting, our platforms enable kitting with many type of heteogeneous objects such as small parts, classic packages or soft material such as clothes.