Design and Implementation of a Constrained Scheduling for an Elevator Group Control System Using the Deep Q Network
碩士 === 國立交通大學 === 資訊科學與工程研究所 === 107 === This thesis proposes a new scheduling method for elevator group control system based on the Deep Q Network with scheduling constraints. To decrease the waiting time of the passengers, we apply the Deep Q Network to assign the hall call request to the specific...
| Main Authors: | , |
|---|---|
| Other Authors: | |
| Format: | Others |
| Language: | en_US |
| Published: |
2019
|
| Online Access: | http://ndltd.ncl.edu.tw/handle/cq7ty4 |
| Summary: | 碩士 === 國立交通大學 === 資訊科學與工程研究所 === 107 === This thesis proposes a new scheduling method for elevator group control system based on the Deep Q Network with scheduling constraints. To decrease the waiting time of the passengers, we apply the Deep Q Network to assign the hall call request to the specific car under the scheduling constraints. Our simulated circumstance is an elevator group control system operated in a 10-stair building with 2 cars for 100 days. The transportation time of the car moving between two adjacent floors is 5 seconds. The door opening time is 11 seconds for passengers getting in or leaving the car. Compared with the Deep Q Network without scheduling constraints, the average waiting time can be reduced by 14.11%. Compared with the greedy algorithm at alpha = 2, the number of average moving steps can be reduced by 19.23%.
|
|---|