Neural Predictive Control of Broiler Chicken Growth

Abstract

Active control of the growth of broiler chickens has potential benefits for farmers in terms of improved production efficiency, as well as for animal welfare in terms of improved leg health. In this work, a differential recurrent neural network (DRNN) was identified from experimental data to represent broiler chicken growth using a recently developed nonlinear system identification algorithm. The DRNN model was then used as the internal model for nonlinear model predicative control (NMPC) to achieve a group of desired growth curves. The experimental results demonstrated that the DRNN model captured the underlying dynamics of the broiler growth process reasonably well. The DRNN based NMPC was able to specify feed intakes in real time so that the broiler weights accurately followed the desired growth curves ranging from −12% to +12% of the standard curve. The overall mean relative error between the desired and achieved broiler weight was 1.8% for the period from day 12 to day 51.

Deep learning in agriculture: A survey

“…Although DL has been associated with computer vision and image analysis (which is also the general case in this survey), we have observed 5 related works where DL-based models have been trained based on field sensory data (Kuwata & Shibasaki, 2015), (Sehgal, et al, 2017) and a combination of static and dynamic environmental variables (Song, et al, 2016), (Demmers T. G., et al, 2010), (Demmers T. G., Cao, Parsons, Gauss, & Wathes, 2012). These papers indicate the potential of DL to be applied in a wide variety of agricultural problems, not only those involving images.…”

“…A particular paper investigating segmentation of root and soil uses images from X-ray tomography (Douarre, Schielein, Frindel, Gerth, & Rousseau, 2016). Moreover, some papers use text data, collected either from repositories (Kuwata & Shibasaki, 2015), (Sehgal, et al, 2017) or field sensors (Song, et al, 2016), (Demmers T. G., et al, 2010), (Demmers T. G., Cao, Parsons, Gauss, & Wathes, 2012). In general, the more complicated the problem to be solved, the more data is required.…”

“…From the rest 9 papers, 2 performed predictions of fruits counted (scalar value as output) (Rahnemoonfar & Sheppard, 2017), (Chen, et al, 2017), 2 identified regions of fruits in the image (multiple bounding boxes) (Bargoti & Underwood, 2016), (Sa, et al, 2016), 2 predicted animal growth (scalar value) (Demmers T. G., et al, 2010), (Demmers T. G., Cao, Parsons, Gauss, & Wathes, 2012), one predicted weather conditions (scalar value) (Sehgal, et al, 2017), one crop yield index (scalar value) (Kuwata & Shibasaki, 2015) and one paper predicted percentage of soil moisture content (scalar value) (Song, et al, 2016).…”

“…From the other 32 papers, different trees were used in training and testing in (Chen, et al, 2017), while different rooms for pigs (Demmers T. G., Cao, Parsons, Gauss, & Wathes, 2012) and chicken (Demmers T. G., et al, 2010) were considered. Moreover, Hall et al…”

“…Particularly interesting were the approaches employing the Faster Region-based CNN + VGG16 model (Bargoti & Underwood, 2016), (Sa, et al, 2016), in order not only to count fruits and vegetables, but also to locate their placement in the image by means of bounding boxes. Similarly, the work in (Dyrmann, Jørgensen, & Midtiby, 2017) Recurrent Unit) (Minh, et al, 2017), estimating the weight of pigs or chickens (DRNN) (Demmers T. G., et al, 2010), (Demmers T. G., Cao, Parsons, Gauss, & Wathes, 2012) and for predicting weather based on previous year's conditions (LSTM) (Sehgal, et al, 2017). RNN-based models offer higher performance, as they can capture the time dimension, which is impossible to be exploited by simple CNN.…”