Robotic arm heralds breakthrough in robot learning
The EU-funded DECORO (Developmental context-driven robot learning) project focused on developmental, context-driven robot learning, which seems key if robots are expected to fulfil their promise in improving lives and living. DECORO set out to better understand the relationships among the ‘state’ used in a learning algorithm, the embodiment of the robot itself and the robot’s sensory context. The DECORO team first created an architecture for exploring real-time learning with high-dimensional contexts. A key finding was that when allowing the robot to take into account a larger set of sensory stimuli, it can better overcome sensory noise and forgetting when performing similar tasks. The scientists then developed an open source, 3D printed robotic arm that can interact physically with the environment while developing its sensorimotor skills. The arm combines structural components that are printable on hobby-grade 3D printers, and rubbery tendons in an agonist-antagonist configuration. This enables easy replication, robustness even to fast impacts, a repair cycle of minutes when something breaks, and stiffness and damping when required. It is able to survive impacts and physical exploration with objects that are not modelled by using variable stiffness actuators with passive compliance. The robotic arm employs simple internal models that relate muscle lengths to joint angles and stiffness. These models are learned through a self-calibrating procedure, and are used to perform fast and targeted movements, as well as collision detection. The arm can reduce the typical oscillations at the endpoint while moving fast through the dynamic use of co-contraction when performing movements. Lastly, the project partners began the work of exploiting the robotic arm in order to develop context-driven behaviours. With open-source software and hardware, DECORO’s soft and robust arm is able to respond uniquely to its environment. It is being replicated in multiple universities across the globe. Another research avenue is for picking tasks in agriculture, and an SME has been established in the United Kingdom for this purpose, Fieldwork Robotics Ltd.
Keywords
Robotic arm, robot, robot learning, developmental learning, DECORO