Soft Robots for Agricultural Purposes
The concept of soft robots is considerably foreign to most, and are still reserved to research labs. For movie lovers, a soft robot probably elicits thoughts of the adorable Baymax. Well, that is not too far from the truth. Soft robots are traditionally made from an elastic material molded into a specific geometry. The robots primary form of actuation is preformed by pneumatic pumps. These pumps combined with the geometric constraints of the elastic body, create specific predictable motions.
As an engineer, of course, I’m interested in application. Orthodox robots are rigid and can be programmed to perform very fine manipulation. Soft robots, while still new, do not have the fine motor skills that can be achieved through programmed robots. What utilitarian functions can we find for soft robots? In fact, soft robotics labs have already found a considerable number of applications. Applications for health and biological purposes are already underway. A soft manipulation structure can be useful when interfacing with living organisms. Developing soft robots is cheap. At a recent robotics conference, I listened to Dr. George Whitsides of Harvard University discuss how arthropod inspired robots can be used as disposable sensor robots in hazardous environments.
A specific application of soft robotics that interested me is agriculture. Most automatic solutions for harvesting only operate on large sturdy crops. There exist few mechanisms for collecting smaller crops, but many are violent and tend to damage the plant in the process. There have been advances by companies such as Soft Robotics Inc, that incorporate soft elements into their robotic grippers. These grippers still depend on the object of interest to have some sturdiness.
Such predicaments led to an undergraduate research project I participated in. The projects goal was to explore the autonomous picking of light fragile berries, specifically raspberries. A unique gripper was desired for that task. Considering the fragility of raspberries, this was an opportune problem to solve with soft robotics.
The generated idea involved creating a tube like structure that would move up to a berry bush and a isolate a raspberry from its neighbors. The end effector would then further slide up the stem of the berry and then inflate its soft robotic sacks. The sacks do not directly grab on to the berry. Instead, the sacks apply a light pressure on the back of the berry that connects to the peduncle of the plant. The entire tube then pulls away, push the berry off its stem. This is similar to how a person pulls of a berry.
The entire solution involved computer assisted design (CAD) modeling, 3D printing, soft robotic construction, and micro-controller programming. Incipient CAD models of the gripper included a perfectly cylindrical end effector that would contain four inflatable silicon sacks. For future prototypes, the end effector was designed with room for multiple cameras. The eventual goal is to integrate computer vision for robot navigation and berry localization. As the design evolved, a mouth for the gripper was extruded to better isolate the berries. Leveraging the advantages of 3D printers, the external pneumatic piping was built into the walls of the end effector. This created a more compact and convenient design.
Of course, generating the silicon soft components, was the most exciting part. With the help of my non-existent chemistry background and crude dorm room lab space, I was able to make surprisingly functioning silicon sacks. The sacks were generated from a two part silicone mixture known as Ecoflex 00–30. A special two part mold was modeled and 3D printed.
The congealing process was quite lengthy. To get a solid sack took roughly five to six hours. Because of my rather inefficient process and rudimentary tools, the sacks did not always form correctly. For example, I did not have a traditional sealing and releasing agent. These are used to form smoother surfaces and make the extraction of the sacks from the mold easier. In substitute, I printed my molds at sixty microns (the highest definition my 3D printer allowed) and used Vaseline as a releasing agent. This produced acceptable results. Be warned however, Vaseline has a tendency to get everywhere and refuse to leave. The sacks had to be thoroughly rinsed after the process.
More sophisticated methods use a vacuum chamber to remove air bubbles that form in the liquid material during mixing. This is highly recommended if one is trying to make anything beyond a simple proof-of-concept. Removing holes from the liquid increases the uniformity of the final product and reduces the possibility of tears. Not using a vacuum chamber is a possible explanation for why some of the generated sacks contained holes and blemishes. Fortunately, Ecoflex sticks very well with solidified Ecoflex. Holes can be filled in if left to sit in the mold with the sack a second time. The hardened silicon can also be easily cut, which is necessary to remove fringes developed along the mold line.
The silicone sacks were attached to the end effector frame with a chemical adhesive and bolted for further support. Future iterations of this design eliminate the use of a glue, but for an initial proof-of-concept the method was quick and functional. In the video below, the grasping and inflation mechanisms are performed with my finger covering an exhaust hole. The final prototype contained a microcontoller controlled servo that was triggered to close the exhaust hole (final image).
I was quite ecstatic when the silicone sacks were successfully integrated and worked. I had considerable apprehension over the durability of the silicone material (again, not a materials engineer) and was concerned that the air bubbles may facilitate tearing. The use of the liquid adhesive was and additional concern. However, even after repeated use, the sacks never broke the seals.
It’s important to remember that this project was a proof-of-concept. The project successfully demonstrated the implementation of the inflatable sacks. Unfortunately, no live tests on berry bushes were performed. Mostly, because the berry of interest was out of season during the time of development. Formal testing of such a device must be conducted.
