We recently wrapped up the first phase of DeMineCraft (DMC) Worlds project at our 3rd virtual open house. In this phase of the project, the focus was to build the first prototype of a sweet potato harvester robot.
Why sweet potato harvester?
The process of removing a landmine is similar to that of harvesting a sweet potato. This makes harvesting sweet potatoes a good milestone to begin with for the DMC project.
Sweet potato is a highly nutritious crop that thrives well in Singapore’s climate. Apart from having high starch/carb content, it is also highly packed with other nutrients like fibre and vitamins. The leaves of the crop are also edible. We thought it would make a good pilot trial to value-add to food sustainability with our project.
At IROS last year, we observed that there was a lack of differentiation in form factors for industrial robots so we thought we could start with developing something different. Our biggest hurdle faced was Where do we draw the lines between designing for effectiveness and for aesthetics?
Harvesting sweet potato might not sound that complicated. For humans, it is almost as simple as using a digging fork to loosen the soil and then pulling the tubers out. However, a sweet potato harvester robot would, in our opinion, have to be trained to move around the farm, remove/navigate through the sweet potato leaves, dig, find where the tubers are, harvest the sweet potatoes, store them and transport them back. Hence, for robots to harvest sweet potatoes effectively to support humans, it is quite complex.
Looking into robots that are being researched for demining, we found that most are specialised robots focusing on one subfunction such as mine detection, localisation & mapping, demining, etc, with most efforts concentrated in mine detection and localisation.
Do we need a single robot that can perform all functions? Taking a step back to look at the overall vision of establishing a virtual farming community supported by robotics and automation, we arrived at our conclusion/concept hypothesis. There are common functions across urban farming robots such as locomotion or storage and transportation. Coincidentally, these functions have been well-researched by experts and academics, and there are well established mechanisms/methodologies to implement these. Hence we can remove those functions from the equation and focus our efforts to contribute in a different category of robots, and possibly utilise more exciting form factors.
So we arrived at building a mothership that would take on the important functions of locomotion around the farm, storage and transportation. To unlock more interesting designs and mechanisms that can be applied in the urban farming context, this mothership can also carry “baby bots” that would have specialised function such as digging or harvesting. This way, the project may be extended to harvest other crops using different combinations of specialised robots.
To illustrate, we hypothesised that we can apply a spherical robot mechanism has an inherent advantage in digging. However, it would be a bit of a stretch to require it to move around the farm independently on its own and carry sweet potatoes. By focusing only on digging, it allows us to apply this mechanism (which is also widely studied) to our series of farming robots.
Similarly, nature has it such that most crawlies that are able to navigate soil well wriggle like a worm or snake. This mechanism of moving through soil can be adapted by a robot with a gripper to harvest sweet potatoes and other similar crops. However, a worm robot might not be able to move faster than a robot on legs across the farm, or carry much payload.
We changed the scope of our project to prototype different specialised robots that will collaborate with one another instead of building one fully functioning sweet potato harvester robot. This allows us to use other interesting mechanisms that have yet to find an application.