RIWO
Riwo specializes in industrial automation and machine and process control. We work on (customized) projects which are prepared in our office and commissioned onsite. Within Riwo, we appreciate an informal and collegial working atmosphere.
We differentiate ourselves by relieving our customers by supporting them in project management, organization and selection of machine controls, be in lead of automation development, but especially by investing in ourselves. Most importantly, our employees make the difference. By combining the aspects mentioned above, our customer relations will turn into partnerships.
ROBOTICS
The main focus of the R&D Robotics division is to develop working robotic proof-of-concepts up to TRL 7. All year round we have multiple vacancies for internships and graduation assignments. This document provides a summary of the current projects and assignments. Together with you we can tailor the assignment to match your learning goals and interests.
PROCESS
All of our student assignments are all related to ongoing R&D projects, so your input may end up in a real product at one of our customers.
At our department we work in with the Agile/SCRUM workflow. At the start of you assignment you will write a problem description in your own words, to see if you clearly understand the goal of the assignment. Then for each 3-week sprint you write a sprint report where you discuss the process, implementation and results of the finished sprint and determine the functional requirements and planning for the upcoming sprint.
At the end of your assignment you will present your achievements to the company in the form of a presentation and demonstration.
In return we offer you:
- Possibility to tailor your own assignment
- A market-based internship fee
- Day-to-day technical supervision
- Working in a collegial and familiar environment
These competences will help you achieve the goal:
- C++
- An independent, curious and creative attitude
- You like coding, but you also want to get some hands-on experience with a real setup
Our projects
Orchard robot
Together with universities and partners in industry we are developing a robotic system for autonomous fruit picking and tree pruning, based on a cobot and 3D vision. Riwo focusses on the robot control, the object detection and the overall integration.
Fruit packaging
Riwo is developing a vision system for precise and high speed fruit quality control and pose estimation. This system is placed in-line with a fruit handling system such as a packaging robot.
Pick&place plants
For a local greenhouse Riwo is developing a pick&place system to pick pots with plants from a moving automatically guided vehicle (AGV) and place them into trays on a conveyor belt.
Waste sorting
Together with the applied university Riwo is developing a waste sorting system for a local recycling company. To do this Riwo is developing a vision system in combination with robotic manipulators to sort various materials from the waste flow.
Autonomous vehicles
For our customers that make agricultural vehicles we want to improve driving in unstructured dirty environments, such as orchards or stables. For situations where guidance systems, such as the Riwo Navigation Antenna in combination with induction wire or RFID tags, are not possible we have to be able to drive based on other sensors. Therefore we would like to investigate sensor fusion, vision and SLAM.
ASSIGNMENTS
We are looking for new students for various assignments starting September 2023. In all assignments you will learn to use C++ and ROS2 (Robotic Operating System). You will also get the chance to test your software with physical test setups. These are some examples of assignments that you may work on, assignments can be tailored to the specific interest of the candidate and may change focus depending on the outcomes of the internship and graduation assignments that are currently running:
Depth-based localisation
Our test AGV currently uses wheel odometry, an IMU and the Riwo Navigation Antenna as sensor inputs. We would like to provide depth-based odometry as sensory input, based on a lidar and on a depth camera. The goal is to investigate localisation in a mapped area based on a 2D/3D pointcloud.
Suggested programs: applied computer science / electrical engineering / mechatronics
SLAM
The environment where our AGV’s drive around is not always properly mapped. Therefore we would like to create our own map based on visual input, such as color camera’s, depth camera’s and lidars. Your goal will be to investigate SLAM (simultaneous localisation and mapping) algorithms using these sensors.
Suggested programs: applied computer science / electrical engineering / mechatronics
Obstacle avoidance
Sometimes unknown obstacles may appear in front of the AGV. The AGV has to handle this. Your first goal is to detect a static obstacle, based on depth information. The second goal is to change the local path to move around the obstacle. The third goal is to replan the global path, if the AGV can not move around the obstacle.
Suggested programs: applied computer science / electrical engineering / mechatronics
Orchard simulation
Currently we only have a very minimal visual dynamic simulation of a robot arm, camera’s, apples and trees to verify the apple picking algorithms. The goal is to set up a proper visual representation of the orchard visuals (bright sunlight, wet apples, etc.), and the orchard dynamics (branch bends when you move against it). A second goal would be to include apple ripeness and apple quality in the simulation, as we need to be able to quantify this in the next stage of the project.
Suggested programs: (applied) computer science / creative media
Infrared pear defect detection
Not all quality defects on a pear are visible with a standard RGB camera. Therefore we would like to investigate which defects are visible with an infrared camera. Your first goal is to select the lighting and filters for the infrared camera that we have at Riwo, based on bandwidths where defects can be detected. The second goal is to make an assessment which defects are visible. If there is time left, you can investigate how to use the monochromatic infrared images as input for our current segmentation nets.
Suggested programs: applied computer science / applied physics / electrical engineering / mechatronics
Scoring neural nets
For both the apple and the pear project, we need to classify the quality of the fruit into different classes. The threshold for these classes will vary per season, so we cannot simply train a classification net on the possible classes. Therefore we would like to investigate neural nets that can give a score between 0% and 100% on a certain aspect. Your goal is to select and train a few of these nets to find out which one works (best). A second goal would be to investigate neural nets that can give multiple output scores at the same time, such as ripeness + quality.
Suggested programs: applied computer science / applied physics