Robotic touch, sense, and learning of delicate vegetables


SummaryDespite technological advances in automation in agriculture, there are still roles which can’t be automated where human labourers are still deployed for tedious manual handling of products in uneasy working conditions. This business model is expected to be unfeasible in some year time, especially under the pressure of cost reduction from retailors, the introduction of the “living wage” and uncertainties in immigration policies of low-wage workers.

The challenge addressed in this project is to explore the use of the state-of-the-art robotics technology and to evaluate the feasibility of this approach in practice. The PI's research group has been developing the use of computer vision and machine learning techniques for robotic manipulators that are able to detect physical objects of various kinds with some degrees of uncertainties. Unlike the controlled factory environment where the conventional robots perform exceptionally well, farming processes involve many uncertainties: working conditions (lighting, temperature, humidity, etc.) are often very different; and every vegetable has different shapes and sizes. These uncertainties cannot be completely controlled, thus they are an ideal target application of our intelligent robotic manipulators.  The variety of crops also poses additional challenges, with retailers having different specifications for their crops. Currently crops are sorted by manual labourers after harvesting. The work conducted during this PhD will explore the practical application of these technologies to harvest delicate vegetables and explore how through using computer vision and other techniques decision making can be automated within the harvesting process.

Project code:
CP 172
01 October 2017 - 30 September 2020
AHDB Horticulture
AHDB sector cost:
Project leader:


CP172 Vegebot Final Report 2020