Capstone-Project-IIT-G

Smart Parking Dynamic Pricing System

A real-time, intelligent pricing engine for smart city parking lots that dynamically adjusts rates based on demand, location-based competition, and environmental context (traffic, vehicle type, etc). This project simulates urban-scale parking lot optimization using streaming data and edge-level intelligence.

---

Overview

Urban congestion and inefficient parking pricing lead to traffic delays, lost revenue, and poor utilization. This system models how cities can use data to drive dynamic pricing for parking in real-time, optimizing:

• Space utilization
• Revenue generation
• Customer fairness

Built using Pathway's stream processing framework, the system processes simulated CSV streams and implements three distinct pricing strategies (models).

---

Tech Stack

Layer	Tools/Frameworks
Language	Python 3.11
Stream Engine	Pathway (for real-time table processing)
Visualization	Panel + Bokeh
IDE/Notebook	Jupyter Notebook / Google Colab
Data Source	Simulated CSV stream (parking_stream.csv)
Architecture Tool	Mermaid.js (diagram below)

---

Architecture Diagram

flowchart TD A[parking_stream.csv (simulated real-time feed)] A --> B[Pathway Streaming Ingest] B --> C[Model 1: Timestamp + Schema Parsing] C --> D[Model 2: Demand-Based Pricing] D --> E[Model 3: Competitive Adjustment] E --> F[Final Table: final_price] F --> G[Live Price Visualization (Panel + Bokeh)]

---

Models Overview & Algorithms

Model 1 — Preprocessing & Timestamp

• Reads parking_stream.csv in simulated real-time
• Applies schema parsing
• Converts string timestamp → datetime
• Output: table with real-time streaming rows

🛠️ Tools: Pathway.replay_csv, dt.strptime

---

Model 2 — Demand-Aware Pricing

Dynamically adjusts price based on real-time indicators:

Formula:

• demand_raw = α × (Occupancy / Capacity) + β × QueueLength − γ × traffic_score + δ × IsSpecialDay + ε × vehicle_score

• demand_norm = min(1, max(0, demand_raw / 5))

• price = base_price × (1 + λ × demand_norm)

Coefficients (Tunable):

Symbol	Factor	Default
α	Occupancy weight	1.0
β	Queue length	0.1
γ	Traffic penalty	0.5
δ	Special day bonus	0.3
ε	Vehicle type score	1.0
λ	Demand scaler	0.5

💡 Algorithm: Feature-weighted scoring → normalization → capped pricing

---

Model 3 — Competition-Aware Adjustment

Incorporates spatial dynamics: adjusts price downwards if nearby competitors are cheaper.

Steps:

1. Cross join all parking lots → lot pairs
2. Compute Haversine distance between locations
3. Filter only pairs where price_right < price_left
4. For each lot, compute avg_price_diff = avg(price_left − price_right)
5. final_price = price − φ × avg_price_diff

Parameters:

Variable	Description	Default
φ (phi)	Competition sensitivity	0.5

Algorithm: Cross join → spatial filter → groupby-reduce → price correction

Final Output: final_price ∈ [min_price, max_price]

---

Project Workflow

1. Ingest simulated data as a stream
2. Parse timestamps and schema (Model 1)
3. Compute demand-weighted price (Model 2)
4. Identify and incorporate cheaper nearby competitors (Model 3)
5. Output dynamic final_price
6. Visualize in real-time with Bokeh + Panel

---

Sample Output

• Time series plot of final_price per lot
• Reacts to spikes in demand, traffic, and competition
• Highly tunable via phi, lambda, alpha, etc.

---

Files Included

File Name	Description
Sample_Notebook.ipynb	Base model development notebook
dataset.csv	Sample CSV used for simulation
README.md	Project documentation

---