Controlling complex industrial systems can be a challenging task as it requires extensive knowledge and skills that are usually acquired through years of experience. This makes it difficult to program such expertise into machine algorithms. In this paper, we present a use case that demonstrates how we built control algorithms for a CNC machine using historical logging of observations from experts. With the advent of digital technologies, machining parts are now controlled by computer programs that offer high precision and speed. However, unforeseen scenarios can still arise, which demand operators’ attention and intervention, even with finely crafted machine programs. For our experiment, we collected data from a 5-axis Mazak Integrex i500-series CNC machine over a month manufacturing multiple instances of the same part. We collected observational states, which are sensor data that match the information operators receive and output engagement feed rates following the operator’s trajectories. Using behavioral cloning, we built an initial control policy from this data, testing three families of machine learning models: regression models, ensemble methods, and deep neural networks. The results showed that ensemble methods outperformed the baseline model significantly, proving that they have learned useful control patterns. The policies also demonstrated that ML models could eliminate noisy behaviors from operators’ actions. We believe that with interactive demonstrations in the future, these models have the potential to fully mature. Overall, our study demonstrates the feasibility of building control algorithms for complex industrial systems using historical expert demonstrations and machine learning techniques.

Short-Term Load Forecasting (STLF) is a fundamental component in the efficient management of power systems, which has been studied intensively over the past 50 years. The emerging development of smart grid technologies is posing new challenges as well as opportunities to STLF. Load data, collected at higher geographical granularity and frequency through thousands of smart meters, allows us to build a more accurate local load forecasting model, which is essential for local optimization of power load through demand side management. With this paper, we show how several existing approaches for STLF are not applicable on local load forecasting, either because of long training time, unstable optimization process, or sensitivity to hyper-parameters. Accordingly, we select five models suitable for local STFL, which can be trained on different time-series with limited intervention from the user. The experiment, which consists of 40 time-series collected at different locations and aggregation levels, revealed that yearly pattern and temperature information are only useful for high aggregation level STLF. On local STLF task, the modified version of double seasonal Holt-Winter proposed in this paper performs relatively well with only 3 months of training data, compared to more complex methods.

In this paper, we employ automatic programming, a relatively unknown evolutionary computation strategy, to improve the non-max suppression step in the popular Canny edge detector. The new version of the algorithm has been tested on a dataset widely used to benchmark edge detection algorithms. The performance has increased by 1.9%, and a pairwise student-t comparison with the original algorithm gives a p-value of 6.45 x 10-9. We show that the changes to the algorithm have made it better at detecting weak edges, without increasing the computational complexity or changing the overall design.

A variety of machine learning techniques have been employed to automatically create control algorithms for autonomous vehicles. Much research has focused on various ¿black box¿ approaches, in which the synthesized or learned control algorithms perform well when tested, but are difficult or impossible to analyze and understand. This paper presents the use of the ADATE system to evolve a control algorithm based on a racing car simulator. The system evolved compact and analyzable yet sophisticated control algorithms capable of driving millions of randomly generated tracks at high speeds without ever driving off the road. The approach presented is likely to be applicable to most automatic control problems, given a set of training examples and a suitable software simulator.