Robots in agriculture: Predictive planning improves fruit harvesting

How can robots and humans work well together? Abhishesh Pal aims to create a system where robots can reduce the physical burden on humans and improve productivity in the agricultural sector.

In his doctoral work, "An End-to-End Learning-based Framework for Anticipatory Scheduling of Logistic Support Robots in Fruit Harvesting Scenarios", Abhishesh Pal has developed a predictive planning framework that is designed to improve and increase the efficiency of fruit picking using robots. He completed his doctorate at the Norwegian University of Life Sciences (NMBU).

"We have developed a predictive planning framework that allows robots to predict and adapt to human activities within agricultural environments," says Pal.

This framework uses global stochastic predictions of the fruit pickers' location and local precise detections using visual sensors, such as cameras and laser scanners. Their experiments showed that the use of RGB-D cameras and 2D laser scanners on both mobile and quadrupedal robots could effectively detect, trace, and classify fruit pickers and their activities, ensuring efficient and safe operation.

Put simply: Robots use algorithms to predict events and create a plan based on these predictions. The stochastic predictions do not give one specific outcome, but distribute possible outcomes based on how likely they are.

"The method of using predictive planning and stochastic predictions enables robots to handle uncertainty and complexity in the environment they operate in," says Pal.

Transformation of agricultural practices

The main goal of Pal's research is to improve collaboration and interaction between humans and robots (Human-Robot Collaboration and Human-Robot Interaction) in the agricultural sector. He aims to develop a system where robots can assist autonomously in fruit picking, reduce the physical burden on humans, and improve overall productivity.

Why is this so important? Agriculture is facing significant challenges, including a shortage of labor and increasing demand for food due to a growing global population. By including robots in daily operations, production can become more efficient and less dependent on human labor.

"Safe and efficient interaction between humans and robots in agriculture can transform how tasks are managed, ensuring more consistent productivity and perhaps also lower costs," says Pal.

Pushing forward the automation of agriculture

Pal's research can have a significant impact both within the field and in a broader context. His doctoral work contributes to the fields of robotics, artificial intelligence, and agricultural engineering by pushing forward the capabilities of predictive planning and human-robot interaction. The work has the potential to change agricultural practices by making agriculture more sustainable and efficient. By reducing the need for human labor in risky or repetitive tasks, it can help address labor shortages and reduce injuries, while setting a precedent for similar adaptations in other sectors.

"This research lays the groundwork for future studies and advances in robot autonomy and cooperative systems," says Pal.

Abhishesh Pal fdefends his PhD thesis "An End-to-End Learning-based Framework for Anticipatory Scheduling of Logistic Support Robots in Fruit Harvesting Scenarios" on 27 september. Read more about the event here.