Developing smarter warehouse robots

Researchers from the Rochester Institute of Technology (RIT) are working to develop smarter industrial robots that are designed to be aware of whether or not they have the right of way in busy aisles and can intelligently avoid obstacles, people and other robots. The system integrates smart technologies like LiDAR sensors, and uses artificial intelligence and neural networks to achieve a clearer view of personal space so that robots can behave safely in a decentralised way.

With supply chain challenges brought on by the pandemic and increased demands for e-commerce, technology can provide the support businesses need to improve productivity, efficiency and safety in a warehouse setting.

“This is one area where robotics and autonomous material handling can help,” said Michael Kuhl, professor of industrial and systems engineering in RIT’s Kate Gleason College of Engineering. “Robots can work longer periods of time — not necessarily to replace jobs, but on some of the manual, non-value-added tasks. It means a change of focus of jobs, with people needed to design and maintain fleets of vehicles and robots.”

Kuhl and the project team received a grant for “Effective and efficient driving for material handling”, a one-year, $300,000 project sponsored by The Raymond Corp. It advances earlier work with the company that established task selection and path planning of individual autonomous mobile robots (AMRs).

New work focuses on advanced avoidance and communication strategies for multiple robots and humans in the warehouse environment.

In warehousing operations, there is often a mix of autonomous and human-operated equipment. Avoidance strategies need to be integrated with task options, path planning and recognition of multiple robots able to communicate with one another in real time, and to recognise humans who also will be interacting in the warehouse space.

“We have information about localisation, the different types of sensors that we use within the warehouse to try to identify where the robots are located and the actual movement of the robot,” Kuhl said. “Can they plan to get from the current location to destination safely and efficiently? They can have a short path, but they still need to avoid other robots and people.”

Using deep neural network strategies (types of machine learning techniques), the system components are trained to make specific, sequenced decisions based on common tasks, but also infrequent or unusual actions that might occur in the warehouse environment.

The team is also studying the communication networks within the warehouse — Wi-Fi and cellular network technology functions — as viable solutions. New standards for cellular technologies permit increased individual cellular communication between individual devices, Kuhl explained.

“In terms of people and vehicles interacting, could we take advantage of the sensors of multiple vehicles moving around the warehouse?” he said. “If a vehicle is coming down one path and it sees a person or another vehicle coming out of an aisle, can they communicate and make a decision about what to do next? Who has the right of way?”

The team has found that robots will be able to react in field experiments at Simcona Electronics Corp.

“We needed the real setting to be able to do this work and to move it forward. They provide an extremely valuable resource for us,” Kuhl said.

Image credit: ©stock.adobe.com/au/Monopoly919