PolyHarvest presents an integrated autonomous robotic system for selective strawberry harvesting in UK polytunnel environments, designed to address labour shortages and the practical challenges of commercial deployment. The system combines deep learning based ripeness detection, canopy aware motion planning, adaptive soft grasping with tactile feedback, and probabilistic yield mapping with route optimisation into a single operational pipeline. Evaluated across multiple commercial farms in Kent, Herefordshire, and Angus over two growing seasons, PolyHarvest achieves reliable harvesting performance with low bruising rates and commercially meaningful throughput under realistic field conditions. The results demonstrate that tightly integrated perception, manipulation, and planning can support practical autonomous harvesting and contribute to improved resilience in UK soft fruit production.
Samuel Mbakara John (Wed,) studied this question.