Autonomous Robotic System for object pick and place in warehouse applications

Robotics and artificial intelligence together provide an opportunity to help automate tasks and offer more decent work, increase the overall efficiency, and reduce the final cost. The Autonomous Robotic System project will show that using a multiple degree-of-freedom robot arm, to reach and pick the classified objects, will at a high level performing system create a higher rate of success than conventional approaches.

Background

Autonomous Robotic System for object pick and place in warehouse applications

The Autonomous Robotic System project addresses the need for robotic manipulation in many industries such as manufacturing, automated waste sorting and warehouse order picking.

Background

The Autonomous Robotic System project addresses the need for robotic manipulation in many industries such as manufacturing, automated waste sorting and warehouse order picking.

Presently, handling and manipulation tasks are, to a great extent, done manually by humans, related to repetitive strain or unsanitary conditions (e.g., waste management), where they handle customised orders of varied vegetables, groceries items or different items for a specific contractor or project. Presenting the warehouses, where the items are picked and placed into order bins with major challenges as the demand for automation is driven by reduction of salary costs, reduction of error-prone expeditions and customer complaints, and finally improved work environment. Also with an increasing market for online trading there is a pressing need for new technological solutions to address these challenges.

Purpose / Vision

Automated warehouse order picking is a field of many challenges and has potential for large growth accelerated by development of robotic technologies. This project helps take the first initial steps and will decrease the barriers for the industry to adopt SoA (State-of-the-art) research and technologies.

The project aims to improve capabilities of robotics within recognition and grasping of objects and to do so this project will look at the technological enablers within robotics, vision systems, tactile sensing, sensor fusion and deep learning in a tight interplay with end-users, integrators, and technology providers.

SoA industrial practice is to usually rely entirely on visual input for object classification and pose estimation through either object CAD information or deep learning. In this project, we propose to:

implement an intelligent system based on deep learning and grasping control techniques to facilitate detection, recognition, and successful manipulation of fragile, flexible, unknown, or otherwise challenging objects with a robot manipulator.
deploy the system using multi-fusion data representation (e.g., visual and tactile data) and benchmarked against SoA in terms of object classification and pick-and-place capabilities.

Expected Results

The outcome of the project will provide new insights into how object classification and robot picking in general can be optimised by deploying new methods of fusing tactile and visual sensory data. This new knowledge can lead to innovative solutions that can be further developed and applied in a short-term future by the robotic solution providers in the robot community and enable new business potentials.

Robotic solution providers and robotic companies dedicated to R&D of e.g., tactile sensors, vision systems, robotic grippers, warehouse automation suppliers, manufacturing solution providers, but also other applications where challenging grasping tasks are present e.g., waste sorting applications, food handling, textile industry, among others - will benefit from the results achieved in the project.

The project results can also be used directly as a foundation for developing further funding applications through national and international project calls that will benefit further academic research and Danish robotic companies.