Proposal : Intelligent Elevator Dispatching System (IEDS)

[kimpro82]

Implementation of Multi-Car Elevator Scheduling Algorithm using SCAN and ETA-based Optimization

1. Overview

The goal of this project is to develop a simulation-ready elevator control algorithm that optimizes passenger flow in a high-rise environment. The system will transition from a basic directional SCAN algorithm to an intelligent dispatching model that minimizes Average Waiting Time (AWT).

2. Objectives

• Minimize Average Waiting Time (AWT) and Average Travel Time (ATT).
• Implement a scalable logic that can handle multiple elevator cars simultaneously.
• Provide a visual or data-driven output to verify the efficiency of the dispatching logic.

3. Technical Requirements & Considerations

We are currently evaluating three potential stacks for this implementation:

• Python: Best for algorithmic complexity and data analysis (using NumPy/Pandas for performance logging).
• Excel VBA: Ideal for quick prototyping within an existing corporate infrastructure and easy data entry/visualization via spreadsheets.
• Vanilla TypeScript: Best for a real-time web-based visual simulation (DOM manipulation for car movement).

4. Functional Requirements

R1: Hall Call Management (External)

• The system must register "Up" or "Down" calls from each floor.
• The dispatcher must assign the "optimal" car based on the Cost Function:
  $$Cost = |CurrentFloor - TargetFloor| + Penalty(DirectionMismatch)$$

R2: Car Call Management (Internal)

• The system must handle multiple destination selections within each car.
• Internal stops must be prioritized based on the current direction of travel (SCAN Logic).

R3: Dispatching Algorithm (The Core)

• Phase 1: Basic SCAN (Collective Control) - Move in one direction until all calls are cleared.
• Phase 2: Estimated Time of Arrival (ETA) Dispatching - Assign the car that can reach the floor the fastest, considering current load and intermediate stops.

5. Proposed Architecture (Logic Flow)

1. State Monitor: Tracks all cars (Position, Direction, Load, Status).
2. Request Queue: A centralized buffer for all pending hall calls.
3. Optimizer: A function that iterates through available cars and calculates the lowest cost for a new request.
4. Executor: Updates the state of each car per "tick" or "time step."

6. Success Metrics

• Throughput: Number of passengers delivered per 100 time steps.
• Energy Efficiency: Total distance traveled by all cars combined.
• Wait Constraint: No single passenger should wait more than $T_{max}$ (to prevent starvation).

7. Definition of Done (DoD)

• Core dispatching logic passes edge-case unit tests (e.g., all calls from top/bottom floors).
• Simulation logs the AWT for at least 50 random passenger requests.
• README updated with instructions on how to run the simulation and interpret results.