J. S. Strøm 1 , G. Zhang 1 * and M. Blanke 2

1. Danish Institute of Agricultural Sciences Research Centre Bygholm Intelligent Sensor for Autonomous Cleaning in livestock building- A challenge in bio-environmental engineering J. S. Strøm1, G. Zhang1* and M. Blanke2 1 Danish Institute of Agricultural Sciences, Research Centre Bygholm, Dept. of Agricultural Engineering, P.O.Box 536, 8700 Horsens, Denmark *Email of corresponding author: guoqiang.zhang@agrsci.dk 2 Technical University of Denmark, Ørsted•DTU, DTU-Automation, building 326, 2800 Lyngby, Denmark

2. Danish Institute of Agricultural Sciences Research Centre Bygholm Background • 2 million m2 surface / week • Manual cleaning • Working environments • Automated cleaning • Results / water consumption • Challenges

3. Danish Institute of Agricultural Sciences Research Centre Bygholm vision and goals • To develop an intelligent sensor system that, combined with an autonomous cleaning system, will carry out the cleaning between batches in livestock buildings.

4. Danish Institute of Agricultural Sciences Research Centre Bygholm Research plan • WP 1: Requirements Capture • WP 2: Unclean and clean surfaces and sensor types assessment • WP 3. Smart interpretation of sensor signals • WP 4: Intelligence features and integration with automatic cleaning equipment • WP 5: Sensor system physical design, construction and functionality • WP 6: Dissemination of results

5. Danish Institute of Agricultural Sciences Research Centre Bygholm WP 1: Requirements Capture To provide background information, specify the success criteria and formulate the functional requirements to an intelligent sensor for autonomous cleaning, including: • Characteristics of residue • Requirements- the robot-sensor interface • Quality attributes of remote sensing operation • Requirements - operator and robotinterface

6. Danish Institute of Agricultural Sciences Research Centre Bygholm WP 2: Unclean and clean surfaces and sensor types assessment To provide knowledge about fouling materials and appropriate sensor technology as a basis for selection of the most promising sensor technology • Evaluation of sensor techniques • Distinguish between clean and fouled surfaces • Surface requirements to give reliable information

7. Danish Institute of Agricultural Sciences Research Centre Bygholm WP 3. Smart interpretation of sensor signals • Evaluation of cleanliness under varying conditions • Training, estimation, and validation of an automated classifier/grader • Reflected/emitted electromagnetic radiation and/or • Enable feature extraction for the physical objects

8. Danish Institute of Agricultural Sciences Research Centre Bygholm WP 4: Intelligence features and integration with AC robot • Central control and coordination system for the intelligent cleaning sensor. • Automatic calibration and self-diagnosis features on-line • Fault tolerant methods to enhance system availability • To use and update the knowledge base

9. Danish Institute of Agricultural Sciences Research Centre Bygholm WP 5: Sensor system physical construction and functionality • Test and validate _laboratory conditions • Various types of pollution on different surfaces with various colours, geometry, roughness and shape • Test and validate _near-practical conditions

10. Danish Institute of Agricultural Sciences Research Centre Bygholm WP 6: Dissemination of results • Refereed publication in ISI journals • Selected conferences & seminars • Web-site publications • Workshop & networks • Research educations

11. Danish Institute of Agricultural Sciences Research Centre Bygholm ISAC team, time & Budget • Team: DIAS, DTU, LU,ALTO, Gerni • Time: Nov.2002 – Dec. 2006 • Budget: 7.4 mil. DDK (1mil.USD)

12. Danish Institute of Agricultural Sciences Research Centre Bygholm DELIMITATIONS • To focus on the cleaning level where the pig producers are in charge of the cleaning and to reduce the risk of infection between batches • To focus on cleaning in finishing pig houses • Surface under 1.5 m from floor

13. Danish Institute of Agricultural Sciences Research Centre Bygholm Detection principles • Spectral signatures of surfaces and fouling materials • Hyperspectral imaging with Acousto-Optic Tunable Filters • Multi-specture image analysis • Surface texture image analysis

14. Danish Institute of Agricultural Sciences Research Centre Bygholm PIG HOUSING • 100 million m2 of pig housing surfaces per year in Denmark • type and colour of the materials • New and old inventory

15. Danish Institute of Agricultural Sciences Research Centre Bygholm Floor

16. Danish Institute of Agricultural Sciences Research Centre Bygholm Pen partitions

17. Danish Institute of Agricultural Sciences Research Centre Bygholm Feeding and watering equipments

18. Plastic/concrete corner with fittings Danish Institute of Agricultural Sciences Research Centre Bygholm

19. Danish Institute of Agricultural Sciences Research Centre Bygholm OPERATIONAL REQUIREMENTS • Speed of operation = 7 m2/min • Mechanical requirements • Forces and vibrations imposed • moisture, stains and dirt, etc • Detection of location • Safety

20. Danish Institute of Agricultural Sciences Research Centre Bygholm FUNCTIONAL REQUIREMENTS(1) The cleanliness sensor shall be able autonomously to: • Identify selected surface types in finishing pig houses • Distinguish solid surface from openings • Distinguish non-clean from clean surface • Specify position and area of non-clean surface • Function reliably in the environment of the empty pig house during cleaning

21. Danish Institute of Agricultural Sciences Research Centre Bygholm FUNCTIONAL REQUIREMENTS(2) The sensor shall be able to interact with a cleaning robot to: • Communicate the position of non-clean areas for guidance of the cleaning process • Perform cleaning at least as resource efficient as manual cleaning to a prescribed standard. Interaction with human operator: • Define surfaces to clean, specify cleaning parameters and report back on the result.

22. Danish Institute of Agricultural Sciences Research Centre Bygholm End