Student project: Obstruction Handling for Robotic Motion Planning

What you will do

The goal of the project is to develop a reinforcement learning (RL) approach for robotic motion planning in complex, cluttered environments on enabling a robot to push or manipulate obstructing objects to access a specific target. This non-prehensile manipulation method will allow the robot to clear pathways without needing to take longer path to avoid (movable) obstacles to reach the target, without damaging permanent obstacles (like thinker branches) along the path. This will make robotic systems feasible, less expensive solutions and making it a valuable tool in agricultural settings where obstructions like leaves are common.

Project Initiation and System Design:

• Define specific requirements and constraints based on agricultural applications (e.g., apple or tomato harvesting). 2)Set up or get familiar with existing  simulation environment (e.g., using ROS and Gazebo) to replicate real-world conditions with plants, leaves, and fruit/pruning clusters. 3) Conduct a literature review on similar robotic motion planning methods in cluttered environments.
• Development of Leaf Identification Algorithms: Design a computer vision model (possibly using deep learning) to classify leaves based on whether they can be safely pushed or must be avoided.
• Direction and Push Trajectory Planning: Develop algorithms for determining the optimal push direction based on the robot’s current position and the target fruit location.
• Reinforcement Learning-Based Trajectory Planning: Implement a reinforcement learning algorithm to plan and execute push trajectories, with adaptability for moving flexible obstacles, and external circumstances.
• Testing and Optimization in Simulation: Conduct extensive tests on the real robot, focusing on accuracy in identifying pushable leaves, optimal push direction, and successful trajectory execution.  

What we do for you 

• We have a challenging problem where you have a lot of freedom to come up with solutions.
• We have a diverse team of experts both from the biological and the technical sides to supervise and support you.
• You will join the Data Science team of OnePlanet, which employs state of the art knowledge on machine learning for precision agriculture and the frameworks necessary to perform these big data tasks at huge scale.
• You will be able to exchange views and knowledge with the OnePlanet and Imec community of experts and scientists, widening your professional network.
• We can help you to improve your coding skills up to industry standards.
• You have access to our cloud solutions to solve this problem allowing you to process large amount of data within reasonable time. 

Who you are

• Knowledge of Python.
• Knowledge of Robotics
• Knowledge of Machine Learning.
• Knowledge of OpenCV, Open3D is a plus
• Experience with ROS and Gazebo is a plus.
• Knowledge of Git.
• Basic understanding of Agile-Scrum.
• You are passionate about bringing positive impact to environmental and societal challenges. 

Interested

Does this position sound like an interesting next step in your career at imec? Don’t hesitate to submit your application by clicking on ‘APPLY NOW’.
Should you have more questions about the job, you can contact Bas Boom (bas.boom@imec.nl).