What it does
UrbanAgents runs a simulation in which virtual residents (agents) who differ in age, income, household size and occupation score candidate sites and choose where to live, and shows the result on a map. Candidate sites are scored on the dimensions 「家賃」, 「通勤」, 「安全」, 「密度」, 「建物」 and 「用途」, and a dimension without data is left out of the scoring. Each virtual resident has its own weights for what it values, and the scored dimensions averaged with those weights give its 「総合満足度」. Each virtual resident chooses the candidate site where its overall satisfaction, plus a small random variation, is highest. In one run, each virtual resident chooses where to live only once. The result of a run shows which dimensions were used in the form 「{active} / {total} 次元で評価」, and data that could not be used is named in the form 「{names}データは未整備のため対象外です。」. The map shows PLATEAU 3D building models for the wards near the position in view, a rectangle marking the area a simulation runs on, and dots where the virtual residents chose to live. The dots are coloured in three tiers by overall satisfaction. For a result run with 「シミュレーション開始」 or opened from 「過去のシミュレーション」, clicking a dot on the map shows that virtual resident's age, occupation and income, its 「総合満足度」, its score on each dimension and its 「意思決定理由」, and 「チャット」 lets you ask that virtual resident questions. In the 「Scenario」 section you choose a measure from 「新駅」, 「公園」, 「用途変更」, 「防災」 and 「再開発」, set 「経度」, 「緯度」 and 「影響範囲」, and compare it with a baseline without the measure. The comparison result shows the average satisfaction for 「ベースライン」 and 「シナリオ」, the difference between them, and 「次元別比較」.
Exports and sharing
The result of a run that finished successfully can be downloaded with 「CSV」 and 「PDF」, and 「共有」 creates a read-only link to it. Clicking 「CSV」 downloads a CSV file that lists, for each virtual resident, its attributes, the longitude and latitude of the chosen site, its score on each dimension, its satisfaction and the text of its decision reasoning. Clicking 「PDF」 downloads a PDF file titled 「UrbanAgents シミュレーションレポート」 that contains 「結果サマリー」, 「エージェント一覧」 and 「スコア分布」. Clicking 「共有」 copies a link that shows the result read-only to the clipboard, and 「共有リンクをコピーしました」 is shown. Anyone who has the link can see, without signing in, the run's name, the number of agents, the average satisfaction and the virtual residents on the map in the shared view. Clicking the button to the right of 「共有」 (hovering over it shows 「共有を解除」) disables the link, and opening that link then shows 「共有シミュレーションが見つかりません。」. 「過去のシミュレーション」 opens results you ran earlier.
Who it is for
It is for municipal and urban-planning work, when you want to try out how a measure such as a new station or a redevelopment shows up in where virtual residents choose to live and how satisfied they are.
What it does not do
A simulation can run only in an area where data for both buildings and zoning is in place. The heading of the ward list shows 「分析可能: {n}区（データ整備済み）」, and a ward shown with 「データなし」 in the list is not covered by simulations. The area that 「シミュレーション開始」 runs on is the whole area where building and zoning data are both present, and its ward names are shown as 「分析対象: {names}（データ整備済みエリア）」. Selecting a ward moves the map, but does not change the area the run covers. A scenario comparison runs on the default area for Shinjuku (longitude 139.68–139.72, latitude 35.68–35.71), whatever ward you selected and whatever 「分析対象」 shows. A scenario comparison runs the baseline and the scenario separately, and the virtual residents are newly generated at random in each run. What is compared is the average satisfaction of the two runs and the average for each dimension. The ranges of the virtual residents' age, income, household size and occupation cannot be set on screen. 「エージェント数」 is chosen in steps of 5, from 5 to 50 in the 「Simulation」 section and from 5 to 25 for a scenario comparison. For a virtual resident whose text could not be written by the language model, 「意思決定理由」 is a fixed template, and the screen does not show the difference. In the runs of a scenario comparison, 「意思決定理由」 is always a fixed template. The dots shown on the map right after a scenario comparison do not open a virtual resident's details when clicked. The screen does not show the source or the year of the data it uses.
Main screens
| Screen | URL | What it does |
|---|---|---|
| Sign-in screen | /login |
Signing in and creating an account, on the 「サインイン」 and 「アカウント作成」 tabs. |
| Email verification screen | /verify-email |
「確認しました」, 「確認メールを再送」 and 「別のアカウントでサインインする」. |
| Password reset screen | /reset-password |
「再設定メールを送信」 sends a reset link by email. |
| Landing page | / |
Shown without signing in; Open takes you to the main screen (/app). |
| Main screen | /app |
The map and the panel on the left (the header, ward selection, 「Simulation」, 「Scenario」, and 「Analytics」 when there is a result). |
| Shared view | /shared/:token |
Shows a shared result as 「共有ビュー · 読み取り専用」, without signing in. |
While you are signed in, the signed-in email address and 「サインアウト」 are shown in the header at the top of the panel on the left.
Clicking 「サインアウト」 returns you to the sign-in screen (/login).
When you sign out, the result and chat that were on screen are cleared from the screen.
Results you ran can be opened from 「過去のシミュレーション」 the next time you sign in.
If signing out fails, no message is shown, the screen does not change, and you stay signed in.
When the header's language button has switched the display to English, 「サインアウト」 is shown as 「Sign out」.
Clicking 「別のアカウントでサインインする」 on the email verification screen signs you out and returns you to the sign-in screen.