Upload 16 files

.gitattributes

@@ -74,3 +74,7 @@ question_answering/Data/traffic_signal_control/planning_QA.json filter=lfs diff=
 question_answering/Data/urban_planning/planning_QA.json filter=lfs diff=lfs merge=lfs -text
 question_answering/Data/urban_planning/reflection_QA.json filter=lfs diff=lfs merge=lfs -text
 question_answering/Data/urban_planning/st_understanding_QA.json filter=lfs diff=lfs merge=lfs -text
+road_planning/Data/india.mg filter=lfs diff=lfs merge=lfs -text
+route_planning/Data/Road_Network/Manhattan_od_0.01.rou.alt.xml filter=lfs diff=lfs merge=lfs -text
+route_planning/Data/Road_Network/Manhattan.net.xml filter=lfs diff=lfs merge=lfs -text
+socio_economic_prediction/Data/Guangzhou.json filter=lfs diff=lfs merge=lfs -text

README.md

@@ -1,3 +1,8 @@
----
-license: mit
----
+---
+license: cc-by-4.0
+task_categories:
+- question-answering
+language:
+- en
+pretty_name: USTBench
+---

road_planning/Data/demo.yaml

@@ -0,0 +1,43 @@
+# reward
+reward_specs:
+  dis_weight: 0.5
+  cost_weight: 0.5
+  build_ration: 0.5
+
+# agent
+agent_specs:
+  batch_stage: false
+
+gamma: 0.9
+tau: 0.0
+state_encoder_specs:
+  state_encoder_hidden_size: [16, 4]
+  gcn_node_dim: 16
+  num_gcn_layers: 2
+  num_edge_fc_layers: 1
+  max_num_nodes: 1000
+  max_num_edges: 3000
+  num_attention_heads: 1
+policy_specs:
+  policy_land_use_head_hidden_size: [32, 1]
+  policy_road_head_hidden_size: [32, 1]
+value_specs:
+  value_head_hidden_size: [32, 32, 1]
+lr: 4.0e-4
+weightdecay: 0.0
+eps: 1.0e-5
+
+value_pred_coef: 0.5
+entropy_coef: 0.01
+clip_epsilon: 0.2
+
+
+max_num_iterations: 100
+#num_episodes_per_iteration: 500
+num_episodes_per_iteration: 1200
+#max_sequence_length: 100
+max_sequence_length: 33
+num_optim_epoch: 4
+#mini_batch_size: 1024
+mini_batch_size: 1024
+save_model_interval: 1

road_planning/Data/india.mg

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:a987c8bebd1da967f45e36591fc6e21a4d7eb0910f1e633be03706cbe53d91eb
+size 117177

road_planning/Data/task_info.json

@@ -0,0 +1,7 @@
+{
+  "task_description": "Design a road network that connects unconnected regions to the existing connected areas in an urban environment. Analyze the provided data to determine which road construction will maximize network connectivity.",
+  "data_schema": "- Connected regions: A list of regions that are already connected.\n- Connectivity: A list of tuples (node_1, node_2, distance) representing existing undirected roads between nodes, where distance is in kilometers.\n- Unconnected regions: A list of regions that are currently isolated from the network.\n- region_nodes: A list of boundary nodes defining the geographical extent of each region.\n- Available roads: A list of potential roads that can be constructed to improve connectivity.\n- road_edge: A tuple (node_1, node_2, distance) representing an undirected road segment available for construction, with the distance specified in kilometers.\n- new_roads: A list of tuples (node_1, node_2, distance) representing additional road segments that may be built following the initial construction.",
+  "domain_knowledge": "- The task is split into two stages.\n- In stage_1, you need to identify the optimal road to construct next that connects the largest number of currently unconnected regions.\n- In stage_2, you need to identify identify the optimal road to construct next minimize travel distances among all regions.\n- Use logical reasoning to determine the best traffic signal activation. DO NOT write any code or algorithms.",
+  "task_target": "determine the most suitable road to build next.",
+  "task_output_type": "decision"
+}

route_planning/Data/Road_Network/Manhattan.net.xml

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:b8e5464e7c2fcb7ad70ccdd3839a0dcc2e58469efae027f1e5781aa1a237a15d
+size 56888478

route_planning/Data/Road_Network/Manhattan.poly.xml

@@ -0,0 +1,21 @@
+<?xml version="1.0" encoding="UTF-8"?>
+
+<!-- generated on 2024-11-18 20:33:53 by Eclipse SUMO polyconvert Version 1.20.0
+<configuration xmlns:xsi="http://www.w3.org/2001/XMLSchema-instance" xsi:noNamespaceSchemaLocation="http://sumo.dlr.de/xsd/polyconvertConfiguration.xsd">
+
+    <input>
+        <net-file value="./Data/Maps/Manhattan.net.xml"/>
+        <osm-files value="./Data/Maps/Manhattan.osm"/>
+    </input>
+
+    <output>
+        <output-file value="./Data/Maps/Manhattan.poly.xml"/>
+    </output>
+
+</configuration>
+-->
+
+<additional xmlns:xsi="http://www.w3.org/2001/XMLSchema-instance" xsi:noNamespaceSchemaLocation="http://sumo.dlr.de/xsd/additional_file.xsd">
+    <location netOffset="4052.54,8033.44" convBoundary="-0.00,-0.00,9319.16,19577.68" origBoundary="-74.018055,40.701552,-73.907618,40.877841" projParameter="+proj=tmerc +lat_0=40.77390187242861 +lon_0=-73.97009843265616"/>
+
+</additional>

route_planning/Data/Road_Network/Manhattan_od_0.01.rou.alt.xml

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:ee1c45fffdb988b7538f7a99ed0e91fee40f2e410560b5ca11b813b23b134882
+size 12960701

route_planning/Data/Road_Network/Manhattan_sumo_config.sumocfg

@@ -0,0 +1,7 @@
+<configuration>
+    <input>
+        <net-file value="Manhattan.net.xml"/>
+        <route-files value="Manhattan_od_0.01.rou.alt.xml"/>
+        <additional-files value="Manhattan.poly.xml"/>
+    </input>
+</configuration>

route_planning/Data/task_info.json

@@ -0,0 +1,7 @@
+{
+  "task_description": "Design a routing strategy to navigate a road network by analyzing traffic conditions and route lengths. For each candidate road, evaluate its congestion level, the shortest route length to the destination, and the congestion levels of its directly connected roads (including 1-hop and 2-hop neighbors).",
+  "data_schema": "- congestion_level: The severity of congestion on a road (higher values indicate worse congestion).\n- shortest_route_length: The shortest distance from the candidate road to the destination.\n- connected_road_congestion_level: The average congestion level of roads directly connected to the candidate road within 1-hop and 2-hop distance.",
+  "domain_knowledge": "- The primary objective is to minimize the vehicle's total travel time.\n- High-congestion roads should generally be avoided, even if they offer shorter routes.\n- You can ONLY select ONE road.",
+  "task_target": "determine the optimal road to enter next to minimize travel time and efficiently reach the destination.",
+  "task_output_type": "decision"
+}

socio_economic_prediction/Data/Guangzhou.json

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:f0180f42547df3a5cdb5638b74141e80e00daa046ff65eae8dbff25f6403a58b
+size 16730934

socio_economic_prediction/Data/task_info.json

@@ -0,0 +1,7 @@
+{
+  "task_description": "You are tasked with predicting urban indicators—specifically, the GDP and population—of Guangzhou for the next years. Historical data for the GDP and population of the target region and its nearby regions will be provided. Additionally, an example region with its GDP and population records for previous years and following years will be included for reference.",
+  "data_schema": "- gdp_in_previous_years: The GDP (in millions of USD) of a region in past years.\n- population_in_previous_years: The population of a region in past years.\n- connectivity: A list of tuples in the format (region_1, region_2, distance), defining undirected edges between regions with the specified distance in m.",
+  "domain_knowledge": "- You need to analyze the time-series data of GDP and population, then discover any potential patterns.\n- You need to consider the spatial correlation among neighboring regions, where closer regions may exhibit similar patterns to the target region.\n- Use your reasoning skill to solve this problem. DO NOT write any code or algorithms.",
+  "task_target": "predict the most likely GDP and population for the target region in the next years.",
+  "task_output_type": "prediction"
+}