With the dawn of artificial intelligence, scheduling software for the construction industry is at a new crossroads and about to change drastically with the incorporation of search technology that makes this possible.
When the consulting firm McKinsey published their article,Imagining Construction's Digital Future" in 2016, the very first paragraph stated aptly: "The construction industry is ripe for disruption. Large projects across asset classes typically take 20 percent longer to finish than scheduled and are up to 80 percent over budget."
Indeed, construction scheduling needs a change. It has always been dependent on a series of events with a beginning and an end, constrained by factors such as labor, time, and equipment.
So how can we leverage new tools such as AI to solve such a dilemma?
For starters, the best mathematical and computational approach to scheduling employs algorithms that use artificial intelligence (AI), namely, a list approach that employs search technology to help optimize a schedule.
Regarding the importance of artificial intelligence, Deloitte characterized it this way: "It’s time for the technology leaders across the board in every industry to discuss how AI can be used to improve quality, speed, functionality, and even drive top-line revenue growth. A confluence of forces has propelled artificial intelligence into the business mainstream. Add it to the growing list of potentially disruptive forces CIOs can introduce into their organizations for commercial benefit." This couldn’t be any more apropos for the construction industry and the way it schedules its work.
In their 2017 Sixth Annual Construction Technology Report, JBKnowledge discovered from their survey of construction professionals that the importance of automation and technology is key to their future success in construction. Specifically, the report states: "If more construction professionals understood the work tasks that automation and Artificial Intelligence technologies can augment and enhance, they might focus less on the tasks they will 'replace.’”
So, what can AI do for construction planners and schedulers? This article will explain a new method that can benefit from the power of search algorithms. By harnessing the horsepower of search algorithms and "learning" all facets of data in its application, artificial intelligence is simply changing the game of optimal construction scheduling. When used strategically, search algorithms are the secret sauce of construction scheduling.
It’s important to distinguish different forms of artificial intelligence. To begin, a category of artificial intelligence is machine learning. This is an important distinction as machine learning is often confused as synonymous with artificial intelligence; it is not. It is actually a subcategory of artificial intelligence.
Within machine learning, different techniques are used to "learn" from data. Such data is classified as labeled or unlabeled. Labeled data combinations are used to learn, and once trained, unlabeled data can be applied and classified. This takes place in either supervised or unsupervised learning. With supervised learning, labeled data “trains,” an AI and various functions are inferred from it to further classify new unlabeled data. With unsupervised learning, unlabeled data is “mined” or classified and explored.
Almost all scheduling data can be considered unstructured as each project is based on a different work breakdown structure. This fact limits the use of machine learning to be used.
Machine learning is a common tactic of artificial intelligence. You’ll find it in many facets of data science and predictive analytics. And that’s precisely the problem, as there is a better way to schedule.
While machine learning is complex and intricate, the best solution for construction scheduling does not employ it at all. Instead, another approach is used: search algorithms.
Search algorithms are essentially game solvers. They compute data and solve a problem space. These algorithms look at different possibilities and constraints and use various methods to find optimal paths and ultimately solve problems.
Think of a game of chess. Each type of piece has different functionality. Then there's the geometry of the board on which the pieces move. Additionally, all the pieces have different constraints according to the allowable rules of the game. The various pieces can move in different directions. At any given time, one move is better than another, and there is ultimately one best move.
Using the chess analogy, search algorithms yield the best move by using techniques to analyze all the available data along with their possibilities and permutations as different decision nodes. Pieces have possible moves to traverse the board according to the constraints of each one.
One of these techniques is known as depth-first. With depth-first, an algorithm explores a single node, travels as far as it can along that node, and then backtracks. Scientists used this method long ago to solve complex mazes; a single node was explored and then backtracked. The cumulative history of those explorations was then pieced together, in a list, to solve a problem.
A second technique of search algorithms is known as breadth-first, the opposite of depth-first. Instead of exploring the distance along a single node and then backtrack, the node is explored along one depth. All the explorations are aggregated to form a list, and then it moves on to the next depth and explores it.
Yet another technique of search algorithm involves heuristics. In this technique, the algorithm ranks different alternatives. By ranking the alternatives in a list, information can be provided to enable an informed decision regarding which branch to follow.