L-CAS at IEEE CASE 2024: Advancing Automation and Applied Robotics

The Lincoln Centre for Autonomous Systems (L-CAS) made a significant impact at the 20th IEEE International Conference on Automation Science and Engineering (CASE 2024) held in Puglia, Italy, from August 28 to September 1, 2024. Researchers from L-CAS presented six papers, showcasing cutting-edge advancements in robotics and automation, with a particular focus on agricultural applications. These contributions underscore L-CAS’s commitment to addressing critical challenges in modern agriculture through innovative technological solutions.

Papers Presented

  1. “Robot-assisted fruit harvesting: a real-world usability study” by Leonardo Guevara, Prabuddhi Wariyapperuma, Hariharan Arunachalam, Juan Vasconez, Marc Hanheide, Elizabeth Sklar
  2. “Runtime Anomaly Monitoring of Human Perception Models for Robotic Systems” by Hariharan Arunachalam, Zhuoling Huang, Marc Hanheide, Leonardo Guevara
  3. “Mass Estimation of Soft Fruit via Oscillatory Plant Dynamics” by Nikolaus Wagner, Grzegorz Cielniak
  4. “AGRIDS: An Advanced Multi-Modal Mapping Architecture for Robotics and Agriculture” by Jonathan Cox, Marc Hanheide, Riccardo Polvara
  5. “Sparse Gaussian Process Regression for Residual Dynamics Learning in Multi-Rotor Aerial Vehicles Control” by Geesara Kulathunga, Marc Hanheide, Alexandr Klimchik
  6. “Non-Destructive Biomass Estimation Based on 3D Reconstruction from a Handheld Camera” by Jaspret Singh, Grzegorz Cielniak

Research Highlights

Guevara et al. tackled the pressing issue of labor shortages in agriculture by investigating robot-assisted fruit harvesting. Their study focused on the crucial aspect of human-robot interaction, exploring how to integrate robotic assistants into harvesting operations effectively while ensuring picker comfort and efficiency. This research provides valuable insights for agri-robotics companies developing harvest-assist solutions.

Concept video of the work presented by Guevara et al.

Arunachalam et al. addressed the challenge of ensuring reliable object detection in outdoor robotic applications. They developed a runtime monitoring framework that can assess the performance of camera-based detection models in real-time, enhancing the safety of human-robot interactions in agricultural settings. This work is particularly relevant for improving the reliability of robotic systems operating in dynamic outdoor environments. See also Hariharan’s GitHub Repository here: https://github.com/hariharan20/ddom

Analysing the oscillations, presented by Wagner and Cielniak

Wagner and Cielniak presented an innovative approach to non-destructive mass estimation of soft fruits. By analysing the oscillatory motion of plants captured through RGB-D video data, they developed a method to estimate individual fruit mass without physical contact. This technique shows promise for improving yield forecasting and selective harvesting systems, addressing key challenges in sustainable food production.

A view of the AGRIDS system

Cox et al. introduced AGRIDS, a novel multi-modal mapping architecture that bridges the gap between robotics and agriculture. By integrating data from various sources, including robotic and static sensors, AGRIDS enables the creation of comprehensive maps crucial for precision agriculture. This open-source framework facilitates more context-aware decision-making in agricultural robotics applications.

Kulathunga et al. focused on improving the control of multi-rotor aerial vehicles in cluttered environments. Their method, based on Sparse Gaussian Process Regression, models residual dynamics between high-level planners and low-level controllers. The approach demonstrated superior performance in trajectory tracking while avoiding obstacles, potentially enhancing the capabilities of drones in agricultural monitoring and inspection tasks.

Singh and Cielniak developed a non-destructive method for estimating above-ground biomass using 3D reconstruction from handheld camera images. Their cost-effective approach shows strong potential for crop growth monitoring, breeding programs, and yield prediction. By eliminating the need for expensive field robots or extensive training data, this method could make advanced crop analysis more accessible to researchers and farmers alike.

Conclusion

The research presented by L-CAS at IEEE CASE 2024 demonstrates a comprehensive approach to solving critical challenges in agriculture through robotics and automation. From improving human-robot collaboration in harvesting to developing advanced sensing and control techniques, these studies contribute significantly to the field of agricultural robotics. As the agriculture sector continues to face pressures from labor shortages and the need for increased productivity, the innovations presented by L-CAS researchers offer promising solutions that could shape the future of farming. The also used the opportunity to link up again with former L-CAS members and the wider community, and see some of the latest technology.