|M.Sc Student||Mark Levin|
|Subject||Design of an Agricultural Modular Task-Based Robot|
|Department||Department of Autonomous Systems and Robotics||Supervisor||Dr. Degani Amir|
According to the UN’s projections, the world’s population will reach more than 9 billion people by the year 2050. With population growth and the shrinking of agriculture cultivation areas, the population’s feeding needs will meet new challenges. Even today, one of the biggest problems in agriculture is the shortage of labor force, as people prefer to work in comfortable jobs rather than in the field.
One possible solution to this problem is the use of automation and robotics, such as manipulators. Past research has shown the advantage of using these robots for specific, niche tasks such as harvesting cucumbers or sweet peppers. However, a significant limitation to this approach is that single, task-oriented robots can do only a dedicated task that they were designed for. Using a robot of this kind leads to poor utilization since the robot is only used during specific periods. On the other hand, a universal robot is often too complicated and most likely too expensive.
In this work, we suggest a new approach to design an optimal, modular re-configurable agriculture manipulator. We suggest that the joints be made of the same basic parts, so the length of the joints can also be easily adjusted as needed. Installation angles, e.g. angles between different joints, can be also changed during the re-configuration. Below, we separate two types of modularity: with, or without changing the joints order. In both types, changes of length and installation angles are allowed.
We present a proof-of-concept of a modular robot and compare its abilities with a standard, fixed-joint manipulator. In addition, we offer a new approach to the optimization of task-specific robot design. The method is based on a two-stage algorithm. The first stage builds a tensor of data based on a specific tree and possible robot configurations, and the second stage optimizes a user-defined cost function. The entire flow is as following:
- Creating a 3-D model of a tree
- Calculating the tensor
- Describing the optimization problem (both cost function and constraints)
- Solving the optimization problem
The advantage of the proposed method is its versatility. A two-stage method simplifies the process of changing the optimization problem, and solving it, without re-computing the tensor, which is the most time-consuming part. This way, the tensor-building stage needs to be performed only once and various optimization problems can be solved based on the calculated tensor.
To present the advantages of using a modular robot, a number of test cases were checked. Two cost functions were considered for each of these cases: minimizing harvesting time under minimum harvested fruit ratio constraint, and maximizing the number of harvested fruit. All of these tests showed the advantage of using a modular robot.
This work is another step toward increasing the use of agriculture robots and manipulators. The presented proof-of-concept and development method can be used both by researchers and industry to increase the manipulator's utilization and revenue, and overcome the cost-effective problem which prevents robots to be widely used in agriculture.