Autonomous Mobile Service Robot with safe and social-aware navigation for construction
While mobile service robots are increasingly being used to assist or work together with workers in the construction industry, their autonomy and trustworthiness are still limited. Instead of always being controlled by a worker or just spontaneously reacting to what they sense, they should be able to move around workers autonomously and smoothly without interfering with their activities or putting them in direct danger. They should also be able to anticipate environmental changes in order to adapt their behaviour proactively and reliably while ensuring safe worker-robot interaction.
Typical control approaches require system and environment models or a large number of samples (data) for control parametre optimisation. However, this becomes impractical when applied to real-world scenarios with unexpected or unknown environments.
To address the problem, the project “Long-term autOnomy For service robots in consTruction (LOFT)“, researchers from the University of Southern Denmark (SDU), Danish Technological Institute (DTI), and Technical University of Denmark (DTU) together with the seven robotics/AI companies Hecto Drone ApS, Buildcode ApS, Desupervised ApS, Lorenz Technology ApS, Capra Robotics ApS, Meili Robots ApS, Robstruct ApS, and the end-user company NCC ApS have successfully applied robotics and AI to develop the LOFT technology for mobile service robots for construction.
The technology enables the robot to perform proactive and smooth obstacle avoidance based on vision and social awareness built on light indication.
Check it out here >>
The LOFT technology integrates:
- Intelligent robot control, realised by bio-inspired neural control with fast online correlation-based learning (developed by SDU).
- Advanced visual perception, based on faster region-based convolutional neural network with domain adaptation (developed by DTU).
- Social-aware navigation, achieved by using light indication coupled with AI detections (developed by DTI).
Professor at SDU Poramate Manoonpong, who led the project over the past year, sees potential in using LOFT technology for mobile service robots:
“Our results show that the LOFT technology can effectively enable a mobile robot to autonomously and quickly learn without system and environment models and can balance various proactive robot behaviours (smooth motion and collision avoidance) for smooth and safe navigation.
Compared with a conventional control approach, our approach with proactive capability leads to (1) a 29% improvement in the smoothness of robot motion in a static environment, (2) trading motion smoothness for up to 88% fewer collisions along with a 15% better success rate in a highly dynamic environment, and (3) robust obstacle avoidance behaviour in a physical mobile robot system with uncertainties. I believe that this technology can serve as a basis for future “cobot” development.” – Professor Poramate Manoonpong.
About Poramate Manoonpong:
Professor, SDU Biorobotics
Poramate has extensive experience in bio-inspired robot control and learning for autonomous mobile robots. One of his aims is to understand how brain-like mechanisms can be realised in artificial agents so they can become more like living creatures in their level of performance. (www.manoonpong.com).
Mail: poma@mmmi.sdu.dk
About this article and the project “Long-term autOnomy For service robots in consTruction (LOFT)”:
The collaboration project was supported by Odense Robotics, which is co-financed by the Danish Agency for Higher Education and Science as well as the Danish Board of Business Development.
This article was provided by Professor Poramate Manoonpong, Embodied AI and Neurorobotics Lab, SDU Biorobotics, the Maersk Mc-Kinney Moller Institute, SDU, and was one of the projects presented at Robot Innovation Summit 2022 in December.
Re-watch Robot Innovation Summit 2022 here: Robot Innovation Summit 2022 – Odense Robotics
