INFORM Software was recently named one of the '20 Most Promising Artificial Intelligence Solution Providers of 2018', a list created by CIOReview, a US technology magazine based in Fremont, CA.
INFORM has been working with Artificial Intelligence (AI) for over two decades, with commercially available products in use since the early 2000s. "More than 20 years ago, we started developing knowledge-based AI systems that were based on the concept of using fuzzy logic and fuzzy reasoning for representing human knowledge. Over the years, we've added Machine Learning as a second area in our AI activities, and the two are now working in parallel together," Dr. Ulrich Dorndorf, CTO, INFORM, commented.
The unique property of INFORM's so-called "Hybrid AI" approach is that it uses both knowledge- and data-driven algorithms. The former are based on mathematical optimisation, Operations Research, and human know-how, while the latter on advanced analytics, machine learning, deep learning, etc.
Hybrid AI helps customers get the best out of both worlds, as leveraging computer algorithms with human expertise yields results significantly superior to both. Data-driven AI techniques can harvest large amounts of information to detect hidden patterns leading to new insights. They also have the ability to learn, i.e., improve the algorithmic decision-making over time.
On the other hand, mathematical optimisation is typically better capable to solve complex planning puzzles. It also provides much faster response to situational changes and process disruptions, elevating the agility of organizations.
"We do not believe in AI running wild," Adrian Weiler, CEO, INFORM, said. He furthered, "We believe that human control on a meta-basis is essential for AI to function properly. It is like auto-pilot in an aircraft; it performs its role very well 90% of the time but depending on the situation, one can take control back from the autonomous system."
"We've been implementing Cyber-Physical Systems incorporating AI, data analytics, robotics, and human machine interfaces at container terminals around the world for several years now," Dr. Eva Savelsberg, Senior Vice President, INFORM's Logistics Division, explained.
She continued, "We're always on the lookout for ways to implement further substantial improvements that better our customer's bottom lines. In 2018 we spent a good deal of time assessing how we could get the most out of implementing Machine Learning in combination with Agile Optimization, and taking advantage of expert knowledge for the terminal industry."
"The findings are conclusive; Machine Learning offers our customers, and the industry, the ability to drive our already substantial efficiency gains even further. Where we're at with Machine Learning, Agile Optimization, and Expert Knowledge is a delta circuit co-evolution of human, math, and machine," Dr. Savelsberg explained further.
"Humans have a very significant role to play in the foreseeable future of any AI system; it is well documented that human/AI paired systems outperform yet their singular counterparts. INFORM's container terminal software is leveraging this phenomenon to develop market leading solutions," she also said.
Dr. Savelsberg predicts, "We're seeing the first steps in the next significant evolution in niche AI solutions, one that will see a tightening of the human/AI partnership. Our next generation optimizers will create a more tightly integrated communication structure between our AI algorithm's decision-making capabilities and the expert operators."
INFORM's Machine Learning assessment project reviewed data from different terminals and found several areas where improvements could be made to parameters that influence the optimization calculations of their add-on optimization modules. These included, among others: dwell time, outbound mode of transport predictions, and predictions around integrated robotic systems.
Dr. Savelsberg explained, "Take for instance container dwell time. Using Machine Learning, patterns for dwell time can be detected and used for more accurate input to AI systems. The same applies for improving our predictions about the mode of transport with which a container will leave the port. This would add to improvements in storage and prepositioning of containers."
"Furthermore, the behavior of connected robotized systems can be better understood. We analysed such systems and gained decisive knowledge on how reality might differ from expectations. And again, we're able to adjust important assumptions within the optimization parameters. In our assessments, we found that these variables could be improved considerably leading to a noticeable overall improvement in resilience and cost reductions," she summed up.
Read more about the company's solutions and how the IT and transport industries collide in the following articles from the 4/2018 issue of the Harbours Review: