Accurate control of a low-cost soft robotic arm for automated strawberry picking

Reference: CTP_FCR_2019_2

Lead academic supervisors: Prof Gerhard Neuman and Dr Khaled Elgeneidy, University of Lincoln
NIAB academic supervisor: Dr Bo Li, NIAB EMR

This student is to be registered with the University of Lincoln.

Background

The UK alone produces 120,000 tons of strawberries a year (British Summer Fruit Association, 2017) and yet is only the 13th largest producer of strawberries in the world against 3 million tons produced globally. The harvesting of this significant strawberry production is currently heavily labour-intensive, employing over 29,000 of seasonal fruit pickers, who are becoming increasingly difficult to recruit due to the unattractive nature of the job, low wages and seasonal demand. Recognising the challenges in recruiting sufficient human pickers, the industry is under significant pressure to consider automating the harvesting of strawberries via novel robotic solutions.

The University of Lincoln is already actively engaging in research activities into the automation of strawberry harvesting with excellent facilities at the Riseholme campus enabling testing the developed robotic solutions in realistic field environments. The current focus is on automating the transportation of strawberries using a fleet of field robots that autonomously navigate to workers and coordinate amongst each other to maximise their utilisation efficiency. In the next stage, these robots will be equipped with robotic arms to also enable automating the picking process of strawberries. However, a key challenge here is the substantial cost of buying commercial robotic arms to be mounted on a fleet of mobile robots. Moreover, commercial arms are typically rigid and can potentially harm the crop in case of inaccuracies of the vision or path planning system. A novel approach to overcome the cost challenge is through the utilisation of the emerging soft robotic technologies to develop a low-cost soft robotic arm. Being soft in nature has the benefit of not only being inherently safe when interacting with the delicate surroundings, but also the superior flexibility of the soft arm enables passive adaptation to uncertainties expected in the complex field environment. The fabrication of soft arms is relatively simple and inexpensive compared to conventional rigid manipulators, resulting in lighter and more compact solutions that can be easily customised in size and morphology based on the application needs. However, a primary challenge in using soft arms is the difficulty in accurately modelling their complex non-linear behaviour due to the hyper-elastic nature of the materials used in fabrication. Hence, accurately positioning a soft arm to reach a target in a cluttered environment is an important research challenge to investigate.

Objectives

The PhD student will work on the development and control of a bespoke soft arm targeted at the strawberry picking application. This is expected to involve embedding of customised flexible sensors to provide additional positional feedback so that accurate closed-loop control can be achieved. The design of the soft arm will consider the potential for delivering picked strawberries to the robot base by utilising an internal passage through the soft arm, to reduce the need for moving the arm back to the base after each pick and hence shortening the picking cycle. The soft arm body can be potentially 3D printed from flexible materials to automate the fabrication process and yield a more consistent output. Machine learning algorithms will be investigated for online learning control for the soft arm, such that variations from the soft materials and fabrication process can be effectively accounted for using experimental data.

 Approaches

  • Design and fabrication of a low-cost soft robotic arm for strawberry picking application, following the state of the art in continuum manipulators. This design process will provide the platform required for the subsequent modelling and control stages. The actuation of the arm will be achieved through the combination of multi-directional bending and extending soft actuator modules that can be potentially 3D printed.
  • Integrating additional bend sensing capability to estimate the arm profile and hence enable a more accurate closed-loop control of the arm’s end effector pose. The flexible sensors choice will depend on the feasibility of integration with the soft body and the sensory feedback quality required.
  • Data-driven modelling of the soft arm that enables learning the continuum arm kinematics using machine learning algorithms. This would require generating experimental data by supplying different input pressure values to each actuation chamber in the soft arm and visually tracking the resulting end effector pose in the workspace. Hence, the derived models will hence account for sources of variations and non-linearity that are otherwise difficult to accurately model analytically.
  • Performance evaluation in reaching target strawberries in a mock-up setup will be evaluated in terms of accuracy, speed, and flexibility. The results can be also benchmarked against the use of conventional rigid arms to evaluate the expected trade-off between cost and performance.

It is known that for soft material robots, the accuracy and speed are usually limited when compared to conventional rigid robots. However, it is anticipated that the combination of data-driven learning methods and closed-loop control based on integrated sensory feedback, would enable enhanced control of the developed soft arm to reach target strawberries. Moreover, the soft arm will be significantly cheaper to fabricate, use a light-weight design and will be more compact compared to rigid alternatives. In addition, the arm will be inherently safe to interact, which means that workers can still coexist around the soft arm when needed. The envisioned soft arm, although targeted for strawberry picking, has the potential to be further developed in the future for a wider range of harvesting applications that benefit from highly flexible and inexpensive robotic arms.

Applying for this studentship

The most important eligibility criterion for this funded studentship is residency:

  1. UK students: If you have been ordinarily resident in the UK for three years you will normally be entitled to apply for a full studentship, covering tuition fees and a maintenance stipend.
  2. EU students: If you have been ordinarily resident in another EU country (outside the UK) for three years you will normally be able to apply for a tuition fees-only award (without a maintenance stipend). If you have lived in the UK for three years you may be eligible for a full studentship.

This eligibility is unaffected by Brexit. The UK Government has guaranteed EU eligibility for Research Council funding for PhDs beginning before the end of the 2019-20 academic year.

Anyone interested should contact the FCR-CTP administrator for an application form and return the form to the FCR-CTP administrator before the deadline of 30th April 2019.