Organizations: School of Engineering and Materials Science, Queen Mary University of London, Mile End Rd, Bethnal Green, London E1 4NS, United Kingdom. · Department of Robotics, Ritsumeikan University, Japan
The global agri-food sector faces increasing challenges from labour shortages, high consumer demand, and supply-chain disruptions, resulting in substantial losses of unharvested produce. Robotic harvesting has emerged as a promising alternative; however, evaluating and training soft grippers for delicate fruits remains difficult due to the highly variable mechanical properties of natural produce. This makes it difficult to establish reliable benchmarks or data-driven control strategies. Existing testing practices rely on large quantities of real fruit to capture this variability, leading to inefficiency, higher costs, and waste. The methodology presented in this work aims to address these limitations by developing tunable soft physical twins that emulate the stiffness characteristics of real fruits at different ripeness levels. A fiber-reinforced pneumatic physical twin of a kiwi fruit was designed and fabricated to replicate the stiffness at different ripeness levels. Experimental results show that the stiffness of the physical twin can be tuned accurately over multiple trials (97.35 - 99.43% accuracy). Gripping tasks with a commercial robotic gripper showed that sensor feedback from the physical twin can reflect the applied gripping forces. Finally, a stress test was performed over 50 cycles showed reliable maintenance of desired stiffness (0.56 - 1.10% error). This work shows promise that robotic physical twins could adjust their stiffness to resemble that of real fruits. This can provide a sustainable, controllable platform for benchmarking and training robotic grippers.
Figures & tables
Fig. 1: Overview of the proposed methodology and potential applications. (a) The developed soft physical twin of a kiwi fruit is being grasped by a commercial robotic gripper. (b) The proposed methodology begins with the mechanical characterization of real fruits to extract stiffness properties, which can be replicated through a robotic physical twin with controllable stiffness and integrated sensor feedback. The developed twin has potential applications for sustainable, repeatable testing to characterize, benchmark, and improve various robotic grippers/manipulators through adjustable stiffness and sensor feedback. The illustration of the Kiwi fruit in (b) and the grippers in (c) was generated using ChatGPT.
Fig. 2: Mechanical characterization of real kiwi fruits across ripening stages. (a) Mean geometric and mass properties of six kiwi samples measured on Day 1 (raw), Day 5 (mid-ripe), and Day 9 (ripe). (b) The force–deformation responses from indentation tests on each day are shown here. Each curve represents one fruit sample indented to 5 mm at 5 mm/min. The experimental setup is also presented here, where a cylindrical indenter is mounted to an Instron 68-TM testing machine. Progressive softening with ripening is evident from the reduction in slope. Damage points on Day 9 indicate early failure in over-ripe fruits compared to other testing days.
Days
Mean Stiffness (N/mm)
STD (N/mm)
1
5.41
± 1.36
5
3.94
± 1.32
9
3.89
± 1.07
TABLE I: Stiffness of Kiwi fruits at 3 mm Deformation
Fig. 3: Fabrication process of the soft kiwi physical twin. Step-by-step fabrication showing (1) mold preparation for casting two silicone pieces, which primarily make up the main body of the physical twin, (2) demolding, (3) Seal preparation to join the two silicone pieces placed in their molds, (4) sealing of both halves securely by placing a mass on top, (5) fiber-reinforcement using crossing nets, and (6,7) final outer-layer sealing to form the complete fruit.
Fig. 4: Hardware setup and system architecture of the pneumatic pressure control platform. (a) Experimental setup showing the physical twin mounted on a modified CNC machine serving as an X–Y stage, integrated with a force/torque sensor, a syringe pump, a pressure sensor, and a 16-bit ADC for data acquisition. (b) Gripping experiment using the Robotiq Hand-E gripper with the developed physical twin secured in a clamp. (c) System schematic of the physical twin illustrating electrical, I²C, serial, and pneumatic interconnections between the syringe pump, sensors, Arduino, and PC used to pressurize the physical twin for stiffness tuning.
Fig. 5: (a) The average pressure data from 5 trials (blue dots) against the measured stiffness is shown here. The red line shows the fitted first-order polynomial. The pressure data standard deviations were negligible. (b) The average measured stiffness across 5 trials is shown here, compared with the desired stiffness the physical twin was tuned to achieve. The black bars represent the standard deviations in the measured stiffness. (c) The MSE and RMSE between the measured and desired stiffnesses are shown here. (d) The average pressure values measured from the physical twin at each gripper compression step are shown here, based on the established stiffness level. The error bars on each point represent the standard deviation in the pressure readings. (e) The force measured with the force/torque sensor during the first and last 5 loading and unloading cycles (refer to legend) is shown here. (f) The internal pressure of the physical twin during the first and last 5 loading and unloading cycles is shown here.
Mohamed bin Zayed University of Artificial Intelligence, Abu Dhabi, UAE · Khalifa University Center for Autonomous Robotic Systems (KUCARS), Khalifa University, Abu Dhabi, UAE · Department of Mechanical Engineering, University of Windsor, ON, Canada