Task Planning for Farm Robots (MSc)
Most robots are restricted to a simple sense-plan-act paradigm of task execution and their potential for performing complex tasks is not exploited. This assignment aims to investigate the application of planning complex robot tasks in the agricultural domain.
Powerful robots of various types are becoming available and affordable for use in the agri-food domain. Their ability to perform tasks autonomously and with no, or little, human intervention saves human labour and enables precise execution of tasks. In the agri-food domain robots do milking, feeding, monitoring, sowing, fertilising, and spraying tasks autonomously. Regardless of how complex the robots may be, most robots follow a simple sense-plan-act paradigm of task execution. The tasks they perform are remarkably routine limiting the application horizon of the robots and creating bottlenecks on their performance.
Automating complex plans using robots is an important area of research and has often been studied using “toy problems”. As the robots become increasingly affordable and versatile, their potential for performing complex tasks in the agri-food domain must also be explored. Performing complex tasks requires techniques for representing complex robot tasks in a more efficient and user-friendly manner. Such techniques can transform existing robots from simple reflex agents into logical agents. In this research we aim at investigating such possibilities using a task planning language and a new software library that is recently released.
The assignment consists of two parts. The first part investigates the state-of-the-art in planning robot tasks in a selected agri-food sector, which could be arable farming, animal farming, or greenhouse farming. The second part aims to demonstrate the feasibility of the “unified planning” library developed by AIplan4EU project through a case study.
AIPlan4EU is EU sponsored research project with the objectives of making planning accessible to users and facilitate the integration of planning ICT technologies. The project aims to make tools for AI planning in diverse applications domains or sectors. One of those application domains or sectors is the agricultural sector.
Planning in its classical sense is about finding a sequence of actions to accomplish a goal. Planning is often made for a fully observable environment. For instance, a route planner in google maps has a complete map as input. Logical AI agents often work in a partially observable environment and can often plan only the next few steps. They discover the next logical actions as more information becomes available through sensors. The knowledge on how to navigate towards the desired goal using the limited information can be specified with PDDL (Planning Domain Definition Language), a language for planning.
While PDDL is a language for planning (in same manner SQL is a language querying databases), PDDL requires a library and a software tool to do the logical reasoning. The unified planning library developed within the AIplan4EU project is such a software library. In fact the unified planning library aims to be accessible and user friendly, which may create a broad opportunity for the future of smart and precision agriculture.
- Investigate how robot tasks are planned in the current state-of-the-art robots used in the agri-food domain.
- Identify complex planning tasks needed in the agri-food domain
- Demonstrate at a conceptual level the utility of planning libraries (such as that developed within AIplan4EU)
- Performing a systematic literature review (SLR) or a broader multi-vocal literature review (MLR).
- Design complex tasks in agri-food domain using PDDL and/or other modelling languages
- Developing a conceptual prototype design or a prototype software within two agricultural robotics case studies
- Kurtser, Polina, and Yael Edan. "Planning the sequence of tasks for harvesting robots." Robotics and Autonomous Systems 131 (2020): 103591. https://doi.org/10.1016/j.robot.2020.103591
- L. C. Santos, F. N. Santos, E. J. Solteiro Pires, A. Valente, P. Costa and S. Magalhães, "Path Planning for ground robots in agriculture: a short review," 2020 IEEE International Conference on Autonomous Robot Systems and Competitions (ICARSC), 2020, pp. 61-66, https://sci-hub.se/10.1109/ICARSC49921.2020.9096177
- Kootbally, Zeid, Craig Schlenoff, Christopher Lawler, Thomas Kramer, and Satyandra K. Gupta. "Towards robust assembly with knowledge representation for the planning domain definition language (PDDL)." Robotics and Computer-Integrated Manufacturing 33 (2015): 42-55.
- Courses: Programming in Python (INF-22306). Optional: Artificial Intelligence (INF-50306, INF-36803) Engineering and Management of Information Systems (INF-31306), Machine Learning (FTE-35306)
- Required skills/knowledge: Programming in Python or Java
Keywords: Digital twins, Robotics, Planning, Farm equipment
- Ayalew Kassahun (firstname.lastname@example.org)
- Bram Veldhuisen (email@example.com)
- Joep Tummers (firstname.lastname@example.org)