Case Study • Summer 2019
Make a self-driving harvester in 2.5 months. HuizingHarvest already had a cheap Chinese harvester, equipped with some sensors. Our part was to integrate it with our FieldBee tractor autosteer system and to add autopilot functionality that we didn’t have before.
We delivered HuizingHarvest software solutions including the Android application, and prototype for autonomous work. Finally, we provided supporting services during a demo event. The harvester can drive in Manual and Autopilot mode, allowing the operator to easily switch between the modes.
in 2.5 months
5 people in
HuizingHarvest is a company that specializes in the international exchange of agricultural knowledge that acts as a service provider in the international agricultural sector, with a focus on HR and technology.
In July 2019, HuizingHarvest sent us a partnership invitation to join them in an innovative project. The task was to make a self-driving harvester that operates in the autonomous mode due to September 13, 2019.
It probably seems impossible to make a self-driving harvester work in such short notice, however, we saw this as a great opportunity and quite an adventurous task. eFarmer team was working really hard and together with the HuizingHarvest team made a self-driving harvester very fast and on time.
Before starting the project, we had a conversation with HuizingHarvest about their needs. We already had a product for navigation and RTK solution with high precision. Our company was planning to build autonomous vehicles, therefore HuizingHarvest partnership seemed like a perfect opportunity to start working on this project.
For building an autonomous vehicle, we had limited resources – a small budget for development. We provided an agile methodology for building this and calculated 750 hours per developer. We estimated what we can do in this short time by clarifying the requirements, roles, and responsibilities with HuizingHarvest.
HuizingHarvest company already had a harvester but they were looking for ways to make it autonomous. After the discussion, we provided them with two cases for the development:
- Long term development solution that would take 2 years;
- The second case is developing a prototype.
According to the needs of this project and the fact that we only had two and a half months, we went for the second case.
HuizingHarvest already developed electric interface and remote controller for the self-driving tractor. They wanted us to develop firmware for the tractor and controller together with interface for setting up mission parameters and controlling the tractor. The harvester must support manual and autonomous mode.
HuizingHarvest shared with us prototype information. We decided to build a small prototype that reproduces all key features of the harvester. Based on all the information, we estimated three stages of the development:
- First prototype;
- Connection of the controller to the harvester;
- Autonomous mission for the harvester.
We had weekly meetings and daily reports to synchronize the work on this project and build a prototype.
The prototype was ready in two weeks and after the first test during the team meeting, we decided to replace the important driver for hydraulics management. This included finding the right solution for upgrading our architecture and integration with the process. After some tests we made improvements to the prototype and then we built the second one.
Problems we encountered
- While testing our small prototype we had a half-a-second delay, but in a full scale harvester the delay was 2-3 seconds. This delay influenced the steering experience.
- User interface for the controller needed improvements in operating some functions of the harvester.
- We worked hard on the calibration of the controller and improving radio transmission. By reducing the size of the transmitted package via radio, and upgrading the hydraulics system we improved latency and steering experience for the harvester.
- During user interface testing in the field, we decided to improve some elements in our application. Brought some functional buttons to the front in the right place. After this, we successfully improved the final product.
Results and conclusions
In three weeks we close the gaps that we have found after the testing, finalized the design of the Android application.
We also found an open-source solution based on the autopilot controller and decided to integrate it with autonomous harvester with an open protocol for integration. It is also easely deliverable to other projects.
We made all necessary installations together with an autonomous controller, RTK, and calibration of sensors.
In only 2 months, with the amazing effort of our great 5 people team we have made a self-driving harvester. As a result, the autonomous harvester project was successfully presented on the 13th September during the HHX Agrobotix Demo event.
FIND OUT HOW WE CAN PUT SOLUTIONS LIKE THIS TO WORK FOR YOU
When we started this project we thought that this long-term solution is mostly applicable only for HuzingHarvest. But on the prototype stage, we decided to work on a unified solution. We specifically selected open-source libraries for simplifying the process of integration with different platforms.
Therefore, this case can be scalable to other companies who would like to upgrade their tractors with autonomous driving solution. This solution is fully scalable to other prototyping stages for similar projects.
Contact us to find out how we can help your company integrate precision farming with your products and services. Become a technology driven company and stay ahead of the competition.