PhD - Safe and legal operation of robots in agricultural environment
| Posting date: | 25 June 2025 |
|---|---|
| Salary: | £35,116 per year |
| Hours: | Full time |
| Closing date: | 25 July 2025 |
| Location: | Newport, Shropshire |
| Remote working: | On-site only |
| Company: | Harper Adams University |
| Job type: | Contract |
| Job reference: | IM-1930 |
Summary
Doctoral Candidate
Safe & legal operation of robots in agricultural environments
Fixed term - 30 September 2028
Agricultural robots and Artificial Intelligence (AI) technologies could soon be helping farmers improve global food security by alleviating labour shortages, increasing efficiency, sustainability and resilience to climate change, and reducing the use of chemicals, fertiliser, water and energy; thereby minimising farming’s environmental impact.
AI machine learning offers a new expedient method of developing control systems for tasks that would be difficult to manage using classical technologies. Agricultural applications present a unique opportunity for AI systems as they often involve repeatable tasks within a relatively low-safety-risk environment, unlike public or transportation applications.
Although some agricultural technology developers are already incorporating AI systems in their products to enhance the control and functionality, there are growing concerns about functional safety regulations and product certification due to the inherent uncertainty of how AI systems make decisions. Classical engineering development guidelines, are difficult to interpret or simply not transferrable to AI systems. There are virtually no satisfactory ways of exhaustively ensuring and demonstrating that these stochastic systems meet the demonstrable, repeatable, and predictable expectations of existing safety legislation. This is hindering their development and delaying their introduction into the market.
The engineering process for ensuring compliance with functional safety requirements involves thorough risk analyses for the intended application and product. This includes evaluating the system’s ability to perform reliably under environmental and technical limitations, and establishing redundancies, where necessary, to ensure these limits are maintained and honoured. Tools like FMEA and HARA are staples in this process which typically involves the establishment of manageable boundaries within which the system is intended to function and demonstration of the system’s safe operation within these regions, usually requiring the system to abort, or go to a default safe state, when the intended conditions are not met.
For example, lane keep assist on automotive vehicles (whereby the vehicle ADAS system attempts to keep the vehicle within road markings through small corrective steering inputs) will immediately abort (or not engage) if the bright white lines that fit a defined and rigid expectation are not clearly visible. These systems use algorithms, rather than AI machine learning, to detect road markings and, if the system does not detect well understood, highly demonstrable and (critically) highly repeatable parameters for lane markings, it gives a null result and does not attempt any corrective steering.
A non-deterministic AI machine learning model for the identical task would not offer this demonstrability or, critically, the repeatability of classical algorithm-based systems. Furthermore, there is no guarantee that a concurrent redundant model will return an identical result for basic cross-checking. To overcome these issues AI system developers often resort to running parallel classical-based systems to act as an oversight of the AI. This is costly, complex, and time consuming, nullifying the benefits of using an AI approach.
This project’s two aims are (1) Establish the best approach for developing machine learning based control systems for agricultural applications that will allow developers to demonstrate that their AI systems meet safety requirements, based on reviewing and interpreting current legislation for non-deterministic AI systems. (2) Create new compliance testing procedures and processes for agricultural AI machine learning systems. These are essential for manufacturers developing these systems, and will accelerating the supply of AI machine learning controlled machinery to farmers unlocking all of the benefits described in the first paragraph.
The secondments planned for this research project are at:
Any European University or Company justified by the candidate that is willing to take them.
Your application should, ideally, respect the AIGreenBots general requirements and eligibility criteria as described her https://aigreenbots.eu/recruitment/general-info
Have a valid European Master’s degree, or equivalent, in law, manufacturing and/or engineering.
Be fluent in legal and technical English and at least one other European language.
Motivation, flexibility, sense of responsibility, ability to listen and compromise, autonomy, and problem-solving skills.
Ability to work with a small cutting edge commercial/academic technology team, and work with a high level of autonomy and integrity when unsupervised.
Candidates should be prepared and able to travel internationally, often at short notice
It is also highly desirable that you have a sound understanding of robotics, machine learning, and AI.
Attractive salary up to £35,116 per annum
Excellent conditions including - social security tax, health costs, PhD tuition fee, mobility allowance, family allowance (if eligible)
Mobility allowance (if applicable): 600€/month
Family allowance (if applicable): 495€/month
Research, training and networking costs covered: Registration and attendance at international conferences.
Please apply online and submit full Curriculum Vitae (to include two referee details) and supporting documents via the Harper Adams e-Recruitment programme at http://jobs.harper-adams.ac.uk by no later than midnight on 15 August 2025
For further information regarding this role please follow this link https://jobs.harper-adams.ac.uk/Vacancy.aspx?id=9624&forced=1
Safe & legal operation of robots in agricultural environments
Fixed term - 30 September 2028
Agricultural robots and Artificial Intelligence (AI) technologies could soon be helping farmers improve global food security by alleviating labour shortages, increasing efficiency, sustainability and resilience to climate change, and reducing the use of chemicals, fertiliser, water and energy; thereby minimising farming’s environmental impact.
AI machine learning offers a new expedient method of developing control systems for tasks that would be difficult to manage using classical technologies. Agricultural applications present a unique opportunity for AI systems as they often involve repeatable tasks within a relatively low-safety-risk environment, unlike public or transportation applications.
Although some agricultural technology developers are already incorporating AI systems in their products to enhance the control and functionality, there are growing concerns about functional safety regulations and product certification due to the inherent uncertainty of how AI systems make decisions. Classical engineering development guidelines, are difficult to interpret or simply not transferrable to AI systems. There are virtually no satisfactory ways of exhaustively ensuring and demonstrating that these stochastic systems meet the demonstrable, repeatable, and predictable expectations of existing safety legislation. This is hindering their development and delaying their introduction into the market.
The engineering process for ensuring compliance with functional safety requirements involves thorough risk analyses for the intended application and product. This includes evaluating the system’s ability to perform reliably under environmental and technical limitations, and establishing redundancies, where necessary, to ensure these limits are maintained and honoured. Tools like FMEA and HARA are staples in this process which typically involves the establishment of manageable boundaries within which the system is intended to function and demonstration of the system’s safe operation within these regions, usually requiring the system to abort, or go to a default safe state, when the intended conditions are not met.
For example, lane keep assist on automotive vehicles (whereby the vehicle ADAS system attempts to keep the vehicle within road markings through small corrective steering inputs) will immediately abort (or not engage) if the bright white lines that fit a defined and rigid expectation are not clearly visible. These systems use algorithms, rather than AI machine learning, to detect road markings and, if the system does not detect well understood, highly demonstrable and (critically) highly repeatable parameters for lane markings, it gives a null result and does not attempt any corrective steering.
A non-deterministic AI machine learning model for the identical task would not offer this demonstrability or, critically, the repeatability of classical algorithm-based systems. Furthermore, there is no guarantee that a concurrent redundant model will return an identical result for basic cross-checking. To overcome these issues AI system developers often resort to running parallel classical-based systems to act as an oversight of the AI. This is costly, complex, and time consuming, nullifying the benefits of using an AI approach.
This project’s two aims are (1) Establish the best approach for developing machine learning based control systems for agricultural applications that will allow developers to demonstrate that their AI systems meet safety requirements, based on reviewing and interpreting current legislation for non-deterministic AI systems. (2) Create new compliance testing procedures and processes for agricultural AI machine learning systems. These are essential for manufacturers developing these systems, and will accelerating the supply of AI machine learning controlled machinery to farmers unlocking all of the benefits described in the first paragraph.
The secondments planned for this research project are at:
Any European University or Company justified by the candidate that is willing to take them.
Your application should, ideally, respect the AIGreenBots general requirements and eligibility criteria as described her https://aigreenbots.eu/recruitment/general-info
Have a valid European Master’s degree, or equivalent, in law, manufacturing and/or engineering.
Be fluent in legal and technical English and at least one other European language.
Motivation, flexibility, sense of responsibility, ability to listen and compromise, autonomy, and problem-solving skills.
Ability to work with a small cutting edge commercial/academic technology team, and work with a high level of autonomy and integrity when unsupervised.
Candidates should be prepared and able to travel internationally, often at short notice
It is also highly desirable that you have a sound understanding of robotics, machine learning, and AI.
Attractive salary up to £35,116 per annum
Excellent conditions including - social security tax, health costs, PhD tuition fee, mobility allowance, family allowance (if eligible)
Mobility allowance (if applicable): 600€/month
Family allowance (if applicable): 495€/month
Research, training and networking costs covered: Registration and attendance at international conferences.
Please apply online and submit full Curriculum Vitae (to include two referee details) and supporting documents via the Harper Adams e-Recruitment programme at http://jobs.harper-adams.ac.uk by no later than midnight on 15 August 2025
For further information regarding this role please follow this link https://jobs.harper-adams.ac.uk/Vacancy.aspx?id=9624&forced=1