Cili-padi picking robot

In addressing the issues of national food sustainability and reducing the dependability of imported basic food supplies, Malaysia needs to start developing a more advanced and reliable system that can improve the overall food production. This includes improving the quality of the seed that can grow faster with better crop and more resilience, improving the pesticide, insecticide, and fertilizer, and advancing the overall cultivation techniques and processes.

In Cili Padi farming, the use of a fertigation system has been evolving with IoT management in terms of fertilizer dripping automation and status monitoring. Cili Padi is a small type of chili that has normally been used in many Asian dishes. With a high quality of seed, planting a Cili Padi using a fertigation system with automated watering dripping directly to the control medium of chili plant, improve overall production and allowing a lot of saving in term of water usage and labor cost. It took about 6 months for the whole cycle of the chili plantation starting from planting 1-month-old seedling transferred from germination tray into the polybag, until the end of the harvesting season.

The use of advanced robotic and automation in the agricultural industry is becoming more common nowadays. Based on the current practice of Cili Padi farming using a fertigation system, the most labor-intensive process is to set up the system and in the harvesting process. Setting up the system is happening once at the beginning of the planting season but, harvesting the chili starting from the second month until the end of the planting cycle. A single trained worker took about 1 hour to pick 1 kg of Cili Padi. This is the main problem statement that we want to address in this project. Developing a robot that can pick up chili crops will drastically reduce the labor cost.

Recently, we had developed an algorithm for Cili Padi detection on 2D image camera vision using the deep learning method. The detection rate is very high with high accuracy too. This initial work opens a big possibility on a chili picking robot. In this study, since the information of the detection need to also include the position of the chili relative to the robot, before it can be used by the robotic arm to pick up the chili accurately, it is important to use a stereo vision camera to extract the depth information of each detected chili. This information is representing an estimated distance and angle of the chili location relative to the camera position. The new optimized algorithm that is inspired by a previously developing algorithm on 2D images needs to be optimized and developed for the stereo vision application. Other than position, using a stereo image can also determine the size of the chili with the color information, we can estimate whether the chili is ready to be picked or not.

Once the position of the chilies relative to the robot arm is determined, the robot arm can be instructed to pick up the crops one by one. Perhaps the speed of the robotic arm may not be able to pass by the human speed, but if we are looking from the perspective of robot durability and endurance that it can work in a long time without taking a rest, will overtake human in term of performance. The design of the robotic arm and mechanism of picking up the fruits is very important to avoid the chili from bruising and degraded the quality of the crops.

The final chili picking mechanism will be assembled in the form of a robot for farmer companion rather than a fully automated robot since we need to make sure that at all times the robot mechanism is fully capable of doing the task with minimal error and with full human monitoring next to robot making it possible to detect the faulty and making the betta testing the robot can be done straight on the field. Validation and testing of the robot performance will be done to analyze the detection processing time and positioning accuracy, chili size and color for maturity prediction accuracy, and finally the mechanism validation of the robotic arm.