I am Dr Sepehr Abrishami, and offsite construction, a field I research, works only if every component arrives in the right order. Here on Reimagining AEC, this episode turns to 4D building information modelling, or 4D BIM, which links the model to the programme so the order can be tested before anyone builds it. Let me start on a site. It is illustrative rather than a real project, but any planner will know it. A contractor is building an eight-storey concrete-framed office block on a tight city-centre plot. There is one access road along the northern edge of the site, a compound in one corner, a concrete core rising in the middle, and two tower cranes. The programme is a bar chart with several hundred activities, and it has been checked. In one fortnight it shows the core pours for the upper floors and the first deliveries of cladding panels. On the bar chart, those two activities sit on different lines and never touch. On the site, they touch. The concrete pump for the core has to stand on the access road, because there is nowhere else for it. The cladding lorries need the same stretch of road to unload. The panels are meant to be stored in a laydown area under the working radius of the second crane, which will be lifting formwork for the core over that area on the same days. None of this is a design error. Every drawing is correct, and the programme logic is correct. The problem is that nobody has looked at space and time together, and the first time anybody will is the morning the lorries arrive. So the question for this episode is this. What does it take to rehearse a construction project before building it, and then to keep checking the real build against that rehearsal once work starts? By the end, you will be able to define 4D BIM and place it within the ISO 19650 framework, explain what it gives a project team and why adoption is still uneven, walk through the six steps that produce a 4D model, and apply it to health and safety and to the live site. We will take this in six parts: what 4D is, what it buys a project, how a 4D model is built, why many teams still do not build one, how it supports health and safety, and how drones, sensors and digital twins carry it onto the live site. Our office block, with its road, its pump and its two cranes, stays with us throughout. Start with what 4D is, and where the idea came from. Linking a three-dimensional computer-aided design model to a construction schedule was conceived in 1986 and 1987, when the engineering firm Bechtel worked with Hitachi to develop one of the first four-dimensional planning tools. Rischmoller and Alarcón record that history in a paper of 2002. The early work was visualisation, geometry attached to a timeline so that the build could be watched as it grew. The principle has not changed. In 4D BIM, each element of the 3D model is linked to an activity in the construction programme, so that the geometry and the schedule share a common reference. A slab is no longer a shape in one file and a bar in another. It becomes one object that knows when it will be built, tied to one activity that knows what it delivers. What has changed is the framework around that link. Under the BS EN ISO 19650 series, and in particular ISO 19650-2 for the delivery phase, a 4D model is not a standalone product. It is a view of the project information model. It is produced, checked, shared and published through the common data environment, or CDE, the managed source of information for the project. In practice, the 4D outputs answer to the same information requirements, pass through the same review gates and need the same approvals as every other deliverable. A simulation that the client relies on for a decision has a status, a revision and an owner. The simplest way to place 4D among the other BIM dimensions is to treat each dimension as a question asked of the same model. 3D asks what the building is and where each part sits, coordinated across the disciplines. 4D asks when each part is built and in what order. 5D asks what it costs, drawing quantities and rates from the same objects to support estimating and cost control. 6D asks how the asset will be operated and maintained, through the information handed over for use. Taken in that order, 4D is the pivot. It gives coordinated geometry a sequence, and a good 4D model is often where quantity take-off, cost loading and handover data begin. On our office block, the 3D model was complete and coordinated. What was missing was the second question. The model knew where every cladding panel would go. Nothing in it knew when the panels would arrive, or what else would be standing on that road. So the first part of the answer is a link: one shared reference between geometry and time, managed under ISO 19650 like any other project information. Now to what the link buys a project. Teams usually give five reasons for investing in 4D: visualisation, communication, clash detection, quantity take-off and site logistics planning. They come down to two principles. The first is that a sequence people can see is a sequence people can agree. The second is that a model linked to time can check itself, in space and in time, before the site does. Start with seeing. Once the activities are linked to the model, the platform can generate a simulation that plays the build in order. The simulation includes temporary works and major plant, so a tower crane goes up, a ground slab is poured bay by bay, and a cladding package moves around the façade grid. A client or funder can follow a three-minute simulation far more easily than a bar chart of several hundred lines. Decisions that would drag across several meetings tend to settle once everyone has watched it. Seeing also changes who joins the conversation, and when. Because the scheme is presented through one shared model, the team can be brought in early, and it typically arrives in two waves. At the strategic and concept stage, the client, the architect and the lead designer set the outline of the project and agree the information requirements. As the design develops, trade contractors and specialists join and add the detailed information that enriches the model, including its 4D views. The earlier the specialists engage, the less rework tends to follow in construction. On our block, the cladding specialist arriving in that second wave is the person who would have said that the panel lorries need the road while the pump is standing on it. The second principle is checking, and clash detection is the best-known example. A BIM manager or coordinator runs the federated model, which combines the discipline models into one, and the software reports where elements from different disciplines occupy the same space. Picture the typical view. The structure is in one colour, the ductwork in another, the electrical containment in a third, and the clashes are highlighted where a steel beam runs through a duct, or where two electrical services compete for one routing void. Structural against mechanical, and mechanical against electrical, are the classic pairs. Each clash is assigned through the CDE to the responsible discipline, so it is logged, tracked and closed out against a named reviewer. In 4D the check gains a second axis. Two objects can be clear of each other in space and still collide in time, because one cannot be installed until the other is in place, or because both need the same crane in the same hour. Quantities are checked in the same way. Because the model is made of classified objects, material types, sizes and counts can be scheduled straight from it. A typical schedule lists each category, such as walls, floors, curtain panels or structural columns, with its family, its count and its material. In 4D those quantities are tied to the activities that install them. Procurement can start earlier and with more confidence, because the quantities reflect what has been modelled, not someone's reading of a drawing. The same quantities then feed 5D cost, provided the elements are classified consistently with the cost structure, so accurate classification pays off more than once. Site logistics is the check that would have saved our fortnight. Temporary works, crane positions, access routes, laydown areas and site offices all compete for space and time, and traditional planning tends to under-represent them. In a 4D model they sit in the virtual site alongside the building. Play the simulation forward on our block and you see the core rising in the middle, the two crane radii sweeping across the plot, and the access road along one edge. You see the pump arrive on the road and the cladding lorries queue behind it. You see the second crane carry formwork over the panel laydown area. Three conflicts invisible on the bar chart appear in a single run, months before mobilisation, when moving a delivery or a laydown area costs almost nothing. So the link buys two things. It makes the plan visible to everyone who must agree it, and it lets the plan test itself against space and time before any of it is built. The third part is how a 4D model is built. It takes six steps, easiest to hold in mind as three stages: the inputs, the link, and the test. The inputs come first. Step one imports the 3D model into the 4D platform from the authoring tool, typically Autodesk Revit or an equivalent. Step two prepares the activity list, including the logical relationships between activities, the rules that say one activity cannot start until another has finished. That list is built in the platform or imported from a planning tool such as Microsoft Project or Primavera P6. At this point you have geometry and logic, still held apart. Then comes the link. Step three estimates the durations, drawing on quantities from the model wherever possible, because model-derived quantities tend to be more reliable than a manual take-off. Step four is the characteristic move of 4D. Each activity is attached to the set of model elements it delivers, so that the schedule drives the simulation. When the activity for the level five slab starts, the level five slab appears. Then comes the test. Step five adds the temporary works, meaning the cranes, scaffolds, hoists and site logistics, as scheduled items in their own right. This step is the easiest to skip, and skipping it is what hid the problem on our site. The permanent works alone play through cleanly. Only when the pump, the lorries and the crane lifts become scheduled objects do they collide. Temporary works dominate site safety and access, so they belong in the programme with the same standing as the structure. Step six is to simulate, review the sequence with the team, and iterate until the programme is buildable. 4D is not a single output. Each review usually sends the planner back to the programme, and that is its purpose. The order carries the logic. Geometry and logic go in, the link joins them, and the temporary works and the review test the result. The fourth part asks why 4D is still not routine, and the evidence makes sober reading. Swallow and Zulu, in a 2019 study in Frontiers in Built Environment, surveyed one hundred and forty-one construction professionals about 4D modelling for site health and safety. About seventy per cent of directors and managers, and seventy-four per cent of other professionals, were aware of 4D. Yet average adoption among the respondents was only about thirty-one per cent. The benefits they recognised were visualisation and clearer communication. The barriers they named were the cost of training, the time to implement and underlying cultural issues. Across the wider literature, five barriers recur, and they come down to three. The first is people. Skills and experience are short, particularly among scheduling staff moving from bar charts to model-linked planning, and resistance to change is strongest in smaller firms and in trades that have delivered well with paper programmes for decades. The second is money and time. Software, training and the hours spent learning are a real cost, and they fall hardest on small subcontractors, who often hold the detailed knowledge a 4D model needs. The third is structure. The procurement route has to let the supply chain in early, and a traditional design-bid-build arrangement brings the contributors and their information in after the programme has been set, too late to shape it. The tools add a further structural problem. Data do not yet pass seamlessly between the authoring environment and the scheduling platform, so the exchange needs careful data mapping and open formats such as IFC, the Industry Foundation Classes. Think back to the two waves of collaboration. A 4D model depends on the specialists of the second wave arriving while the programme can still change, and every one of these barriers works against that. On our block, under design-bid-build, the cladding specialist would have been appointed after the programme was fixed, and the conflict on the road would have stayed hidden until the lorries arrived. So the barriers have less to do with technology than with whether the right people, with the right skills, are in the room early enough. The fifth part is health and safety, where Swallow and Zulu found something else worth holding on to. Their respondents did not see health and safety as the primary purpose of 4D, yet the study identified a clear need for it there. The reasoning is direct. A safe programme starts from a clear understanding of what is being delivered, when, and by what method. A hazard is a condition in a particular place at a particular time. A model that holds place and time together can find that condition before anyone is exposed to it. At the planning stage, the simulation identifies the construction method, the temporary safety facilities and the hazard zones, so all parties agree them at the planning table instead of negotiating them on site. It also picks out the high-risk activities and the controls each one requires. On our block, the formwork lifts over the laydown area are a high-risk activity. The 4D model shows when they happen, so the exclusion zone under the crane can be scheduled for those hours, and the panel deliveries moved out of them. The same model then becomes a way of communicating risk. For inductions and training, an animated walkthrough of a task lets workers rehearse the sequence before they carry it out. That matters most for unusual work, confined spaces and heavy lifts. For risk zoning, areas of the model are colour-coded. A typical example is an access walkway beside the works, with red cones marking a restricted zone where nobody enters, and yellow cones further along marking a cautioned zone where extra care applies. Safety signage can be placed in the model as well, which gives the team one source of truth for what the site should look like, instead of a drawing that says one thing while the site shows another. The test of all this is whether the information in the model reaches the people who need it. This is information management, and the standards now treat it as such. BS EN ISO 19650-6, published in 2025, sets out a framework for identifying, recording, using and sharing health and safety information digitally across the life cycle of an asset. It supersedes the earlier PAS 1192-6. Safety information drawn from a 4D model therefore belongs in the same managed process as the rest of the project information. So 4D finds hazards in the plan, shows workers where they are, and keeps that information under the same control as everything else. Up to this point, the model has told us what we expect. The last part of the answer reverses the direction and asks the site what is happening, and whether it matches the plan. Drones come first. Unmanned aerial vehicles, or UAVs, are used on construction sites for aerial survey, progress capture and inspection, as the critical review I co-authored with Elghaish and colleagues in 2021 sets out. In a 4D workflow, drone imagery is registered against the model, so the as-built state can be compared with the planned sequence. That comparison supports earned value analysis, delay analysis and audit, and it removes much of the subjective judgement behind traditional progress reports. Image analysis adds a safety layer. Kumar and colleagues, in 2024, used drone imagery and a deep learning model to detect whether workers on site were wearing safety helmets, and to raise an alert when they were not. A 2025 review by Sajid and colleagues found drones combined with artificial intelligence and the Internet of Things being applied to site layout, logistics, progress and quality monitoring, safety management and asset management. The Internet of Things, or IoT, brings the same feedback down to the ground. Tags and sensors on workers, plant and materials report their position and status, and environmental sensors track conditions such as dust and noise. In tunnelling and underground works, where the environment is itself a hazard and visibility is poor, IoT supports automated warnings and safety-barrier logic. When these data are tied to the model, risks can be flagged automatically against rules held in it, and continuous data on resource use support lean delivery and carbon reporting. At the level of a single worker, the architecture looks like this. Wu, Wu and Yuce, in 2019, described a wearable system designed with construction workers in mind. A safe node sits on top of the helmet and measures the surroundings: temperature, humidity, ultraviolet exposure and carbon dioxide. A health node is attached to the chest and measures heart rate and body temperature. The health node passes its readings to the helmet over Bluetooth Low Energy. The helmet then sends the combined data over a long-range radio link, known as LoRa, to a gateway on site, which stores them and forwards them to the cloud. So the chain runs from body to helmet to gateway, and on to the information model, where readings can be judged against safe limits. The digital twin brings these data together. A digital twin is a live, data-connected representation of the asset, updated continuously through sensors, surveys and inspections. On a construction project, the 4D model is its natural foundation, because it already holds the link between geometry and schedule that the twin needs as its reference. Research is now combining 4D with computer vision, drone imagery and laser-scan progress capture, so that the model updates itself as the building goes up. The direction of travel is predictive control. Instead of reporting where the project is, the twin forecasts where it is about to be, and flags schedule deviation, cost overrun and safety risk while there is still time to act. Return to our block one last time. Some months in, the drone survey shows the core a floor behind the plan. The twin carries that delay forward and reports that the late pours now overlap the rescheduled cladding deliveries. The planner sees the conflict on the road coming back, and sees it before the lorries do. So the rehearsal continues once work starts, and the site keeps reporting back to the plan. Let me draw the episode together. We asked what it takes to rehearse a construction project before building it, and to keep checking the real build against that rehearsal. It takes a link between each model element and the activity that delivers it, managed through the CDE under ISO 19650. It takes a disciplined build of that link, with temporary works included and specialists involved early enough to shape it. And it takes feedback from drones, sensors and digital twins that warn when the site and the rehearsal diverge. Replay our office block with that in place. The cladding specialist joins while the programme can still move. The simulation shows the pump, the lorries and the crane lifts meeting on the access road, so deliveries move, the laydown area moves out of the crane radius, and the exclusion zone is planned into the hours of the lifts. When the core slips, the twin flags the conflict returning. The site and the people are the same. This time, space and time were looked at together. I would ask you to take three things from this episode. First, judge a programme by whether it has been tested in space as well as in time, and treat temporary works as scheduled objects, because that is where the hidden conflicts sit. Second, when you set up a 4D workflow, start with procurement and people rather than software, because the model is only as good as the specialists who feed it and the point at which they arrive. Third, treat 4D outputs as managed information under ISO 19650, including the health and safety information covered by Part 6, so that the simulation everyone relies on has a status and an owner. For further reading, start with ISO 19650 Parts 1 and 2, and Part 6 for health and safety information. Swallow and Zulu's 2019 paper is the best starting point on benefits and barriers. The episode introducing BIM and the ISO 19650 framework sets out the full family of BIM dimensions that 4D belongs to. Here is the exercise I, Dr Sepehr Abrishami, would set on Reimagining AEC, for the projects you work on or study. Open a programme from one of them, find its busiest fortnight, and ask what stands on each part of the site on each of those days. If the programme cannot answer, you have found the case for 4D.