Harnessing the potential of AI for housing renovation

Arti­fi­cial intel­li­gence (AI) is increa­sin­gly used in reno­va­tion pro­jects. As such, it reminds us that reno­va­tion is an area of inno­va­tion for which we can dis­tin­guish four sepa­rate areas. 

The first is to clas­si­fy to prio­ri­tise. AI is inte­gra­ted into large-mesh diag­nos­tic tools. For example, it will make it pos­sible to iden­ti­fy – based on ener­gy bills in rela­tion to their size – the 1,000 or 10,000 public buil­dings out of 500,000, which will have the grea­test impact per euro inves­ted in terms of redu­cing CO2 emis­sions or ener­gy consumption. 

The second area is the crea­tion of opti­mi­sed reno­va­tion plans, ups­tream of a buil­ding site : they make it pos­sible to improve the layout of a flat, the reno­va­tion of a buil­ding or the rene­wal of an urban area. The third area, at the moment of imple­men­ta­tion, is the cen­sus. Com­pu­ter visua­li­sa­tion tech­niques, which rely on images from various sources (scans, drones etc.), make it pos­sible to take stock of eve­ry­thing on a site, but also to deter­mine the per­cen­tage of pro­gress or to iden­ti­fy risks.

The fourth area is eve­ry­thing that hap­pens when the buil­ding exists : pre­dic­tive main­te­nance and intel­li­gent mana­ge­ment. Based on various sen­sors, AI makes it pos­sible to anti­ci­pate break­downs or to bet­ter unders­tand the pre­fe­rences of the buil­ding’s users. Some smart buil­ding sys­tems go fur­ther by allo­wing inter­ac­tion bet­ween users (or inha­bi­tants) of the buil­ding with conver­sa­tio­nal agents, or by making pre­dic­tions about the needs of resi­dents and thus gene­ra­ting pro­po­sals for per­so­na­li­sed services.

In terms of reno­va­tion, it is cur­rent­ly the second area, desi­gn, that is seeing the most pro­mi­sing advances. 

At the buil­ding level, tech­ni­cal stan­dards are mas­te­red by pro­fes­sio­nals and the contri­bu­tion of AI is not yet major. Howe­ver, it will allow for ener­gy per­for­mance, envi­ron­men­tal impact, or cli­mate resi­lience simu­la­tions accor­ding to various objec­tives and constraints. This will contri­bute to opti­mi­sing reno­va­tion solu­tions from an eco­no­mic and envi­ron­men­tal point of view. In a chan­ging envi­ron­ment, where ener­gy and mate­rial prices are evol­ving rapid­ly as they are now, the use of simu­la­tion allows us to reo­pen the field of pos­si­bi­li­ties and to find ways beyond pro­fes­sio­nals’ cur­rent habits. More gene­ral­ly, auto­ma­tion also allows work to be car­ried out more quickly.

On a neigh­bou­rhood scale (more rare­ly on the scale of an entire city), where things become more com­pli­ca­ted, AI plays a more impor­tant role. Both for ques­tions of resources : the indus­trial trades invol­ved are prac­ti­sed by power­ful eco­no­mic players, often at the fore­front of tech­no­lo­gy ; and for ques­tions of needs : it is main­ly at this scale that new variables appear, which were not pre­vious­ly inte­gra­ted and have no obvious “tech­ni­cal solu­tions”, trans­po­sed into practices.

These issues include the envi­ron­men­tal impact of reno­va­tion acti­vi­ties and the resi­lience and adap­ta­tion of infra­struc­tures to cli­mate change. They are added to the more tra­di­tio­nal ones that have gover­ned deci­sion-making until now : cost of the work, com­fort, ener­gy consump­tion in the eco­no­mic sense (pur­cha­sing power of hou­se­holds, costs for mana­gers), ener­gy consump­tion in the eco­lo­gi­cal sense (emis­sions). A reno­va­tion plan is now a very com­plex matter !

Basi­cal­ly, we are in the busi­ness of deci­sion sup­port. Let’s take an example. In France and in Europe, many reno­va­tion ope­ra­tions concern neigh­bou­rhoods or buil­dings belon­ging to the ‘social’ stock, built for the most part in the thir­ty years fol­lo­wing the end of the last World War. These are the­re­fore col­lec­tive deci­sions, invol­ving public actors and, of course, public money. The pro­cesses are cum­ber­some, the deci­sion-makers nume­rous and the stakes high, because the phy­si­cal dete­rio­ra­tion of these neigh­bou­rhoods goes hand in hand with social and some­times poli­ti­cal pro­blems. Moreo­ver, they are most often heat loss areas, and the poor qua­li­ty of the buil­dings exposes the inha­bi­tants to ener­gy shocks (increase in price, dis­com­fort). Final­ly, the desi­gn of the neigh­bou­rhoods and buil­dings makes them vul­ne­rable to the conse­quences of cli­mate change, par­ti­cu­lar­ly heat waves and floo­ding. There are car parks in these neigh­bou­rhoods that are heat pockets, and those on the banks of rivers may be sub­ject to a flood risk that was negli­gible when they were built.

Reno­va­tion is both an urgent and com­plex sub­ject, which raises new issues that are not easy for public deci­sion-makers to master.

To sum up, reno­va­tion is both an urgent and com­plex sub­ject, which raises new issues that are not easy for public deci­sion-makers to mas­ter. One of the essen­tial chal­lenges is the­re­fore to acce­le­rate and opti­mise the col­lec­tive deci­sion. This is pre­ci­se­ly what AI allows.

How can it do so ? By allo­wing, during the ear­ly stages of the dis­cus­sion, us to auto­ma­ti­cal­ly gene­rate opti­mi­sed repre­sen­ta­tions, plans for example, which are not “the” solu­tion, but which allow the dis­cus­sion to pro­gress. These repre­sen­ta­tions are what is known in mana­ge­ment science as “boun­da­ry objects”: they are suf­fi­cient­ly concrete to consti­tute an inter­face bet­ween social worlds and actors with dif­ferent perspectives.

In a deci­sion-making pro­cess that lasts a dozen months, if we take a worst-case sce­na­rio, AI could save one or two months. This is signi­fi­cant, without being a radi­cal uphea­val. We can also assume that the qua­li­ty of the deci­sion will be bet­ter : it is made in a way that allows eve­ryone to unders­tand the rea­so­ning of others, and on the basis of repre­sen­ta­tions that are easier to dis­cuss than columns of figures. AI, by making it easy to gene­rate pro­po­sals, is a faci­li­ta­tor and opti­mi­ser of human decision-making.

Not in the sense that a com­pu­ter would speak at the table… but a dis­cus­sion dimen­sion can be inte­gra­ted into the desi­gn work that involves AI. The area that inter­ests us in the Reno­vAIte R&D pro­gramme on which I am wor­king is what is cal­led Adver­so­rial Resi­lience Lear­ning (ARL), a concept that enables arti­fi­cial neu­ral net­works to be model­led and trai­ned by put­ting them in com­pe­ti­tive situa­tions. We take two vir­tual enti­ties, which will lite­ral­ly throw them­selves against each other. One will play the role of defen­der ; it will be a model of the urban com­plex based on BIM (Buil­ding Infor­ma­tion Mode­ling : digi­tal models that do not represent geo­me­tric shapes, but objects) and other known ele­ments. The other will play the role of the atta­cker : this can be any cri­sis (cold, heat, floo­ding, with risks on com­fort, on hea­ting sys­tems, risks of cracks or impre­gna­tion, bud­ge­ta­ry constraints, etc.). After seve­ral hours, this com­pe­ti­tion will allow each of the enti­ties to deve­lop the best stra­te­gies and come up with a set of opti­mi­sed urban rene­wal plan pro­po­sals for decision-makers. 

It is alrea­dy wor­king : this concept has been used in Ger­ma­ny in the desi­gn of smart grids, to iden­ti­fy mal­func­tions on the net­work and gene­rate the appro­priate auto­ma­tic responses.

Interview by Richard Robert