10024275Multi-functional robot control platform, executing multi-stage process files in complex, non-deterministic environments, achieving 99.8% performance level reliability in accordance to guaranteed functional outcome predefined by sensoric dataClosedCollaborative R&DInnovate UKThrough our successful Innovate UK Smart project (Application Number: 23797), we completed an initial product demonstrator of the industry-leading Moley Robotic Kitchen and through our existing live Innovate UK Smart project (Application number: 72673) we are building on this technology by redesigning and upgrading firmware and software to develop our Robotic Hand, utilising Moley's novel patented concept of minimanipulations. This intelligent Robotic Hand will be capable of automatically adjusting to changes inside the cooking environment in real-time and perform over 1000 repeatable operations with a 99% optimal performance value. Through these endeavours we have identified the shortcomings of current robotic training processes. Current technologies are costly in terms of execution time, computational load, sensory hardware and data-accuracy and result in a sensory and computational error-accumulation that cannot guarantee a trajectory solution at every time-step. Similarly, our existing robot planning/training system is based on continuous and iterative multiple-level adjustments of robotic apparatus parameters at each time-step. This approach makes these technologies unscalable for commercial rollout and industry-wide application. The aim of this project is to build on Moley's existing robotic software architecture and functionality in order to deliver a robot control software package that serves the industry's need for robotic manipulations in highly interactive, dynamic and non-deterministic environments that maintains the requested time-parameters fit to the regional process file without requiring continuous sensing and re-planning at every time-step. The proposed solution significantly lowers the sensory and hardware requirements of apparatuses utilising this technology, thus allowing for scalability into more complex environments. Additionally, this technology decreases the long re-training/-calibration procedures by 90%; procedures that are required when reprogramming a robotic apparatus for different operations. We plan to develop our patented, multi-level minimanipulation training software and showcase it on a robotic apparatus that executes trajectories in a complex, non-deterministic environment. Successful completion of this project has the potential to revolutionise the applicability of robotics in commercial kitchens found in industries such as hospitality, retail and healthcare, and will broaden the application of this technology into otherwise unattainable markets such as chemical laboratory equipment and pharmaceutical production.