Demo Abstract: Collaborative Real-Time Scheduling (CRTS)
Algorithm for AGV Transportation System within a CPS
Architecture
Himansu Shaw
University of Houston
Houston, USA
hshaw@cougarnet.uh.edu
Albert M. K. Cheng
University of Houston
Houston, USA
cheng@cs.uh.edu
ABSTRACT
The use of Autonomous Guided Vehicles (AGVs) is important in
smart factories to enhance manufacturing operations. However, cur-
rent AGV scheduling algorithms are inadequate and not connected
to real-time data on manufacturing needs, leading to underutilized
AGVs and missed delivery deadlines. The lack of algorithms that
ensure timely delivery of materials results in idle time of AGVs and
material shortages. We present a collaborative real-time scheduling
(CRTS) algorithm for AGVs in smart factories. The algorithm not
only ensures timely delivery of materials to processing units but
also predicts the minimum number of AGVs required. The algorithm
is designed to operate within a cyber-physical system architecture,
where AGVs and processing units exchange data via a wireless net-
work. The simulation results on the Node-Red platform show that
the algorithm is efficient and adequate to meet real-time delivery
requirements, with an average AGV utilization of over 92%.
ACM Reference Format:
Himansu Shaw and Albert M. K. Cheng. 2023. Demo Abstract: Collaborative
Real-Time Scheduling (CRTS) Algorithm for AGV Transportation System
within a CPS Architecture. In International Conference on Internet-of-Things
Design and Implementation (IoTDI ’23), May 09–12, 2023, San Antonio, TX,
USA. ACM, New York, NY, USA, 2 pages. https://doi.org/10.1145/3576842.
3589177
1 INTRODUCTION
While fleet manager-based AGV systems have been used in mod-
ern industries for 50 years, these systems only prioritize collision-
free path planning [1] and localization [2]. They do not consider
real-time operational data to improve scheduling or optimize AGV
utilization. The paper [3] proposes a real-time scheduling method
for battery assembly using a single processing unit and an AGV
transporting one type of part. A smart AGV manager server re-
ceives battery count and travel-related information to assign de-
livery tasks using a dynamic threshold approach. However, actual
manufacturing scenarios employ a fleet of AGVs for transporting
various parts to different locations.We have expanded this approach
Permission to make digital or hard copies of all or part of this work for personal or
classroom use is granted without fee provided that copies are not made or distributed
for profit or commercial advantage and that copies bear this notice and the full citation
on the first page. Copyrights for components of this work owned by others than the
author(s) must be honored. Abstracting with credit is permitted. To copy otherwise, or
republish, to post on servers or to redistribute to lists, requires prior specific permission
and/or a fee. Request permissions from permissions@acm.org.
IoTDI ’23, May 09–12, 2023, San Antonio, TX, USA
© 2023 Copyright held by the owner/author(s). Publication rights licensed to ACM.
ACM ISBN 979-8-4007-0037-8/23/05. . . $15.00
https://doi.org/10.1145/3576842.3589177
by developing a Collaborative Real-Time Scheduling (CRTS) algo-
rithm incorporating multiple AGVs, processing units, and parts
without a centralized server for information processing.
2 PROBLEM STATEMENT
2.1 Shopfloor Transportation System
We simulated a simplified manufacturing shop floor using a matrix
design with 4 processing units (PUs) and 5 part stores. The AGVs
are identical, and there is a designated AGV home position where
idle AGVs relocate. The layout of the shop floor is shown in Figure
1. Each PU is processing a specific type of parts [A-E]. Each part is
processed at different consumption rate. CR
i,j
is the consumption
rate (number of parts per minute) of part
j
by PU
i
(PU ID i=1 for the
PU
1
, i=2 for PU
2
and so on. part ID j=1 for part type A, j=2 for part
type B and so on.). Each of the parts (A-E) are physically different
in terms of size and weight so the payload capacity of the AGV
for each part is different. For example a heavy part the AGV can
carry fewer number of parts per delivery and for lighter part it can
carry more part per delivery. We can define the payload capacity
of AGV by the term PC
j
(payload capacity of AGV for part
j
. At
any point of manufacturing operation, all the AGVs can be at any
location of the shop-floor. So delivery time by each AGV of part
j
to
PU
i
is also different. We can define the delivery time by the term
DT
k,i,j
delivery time by AGV
k
of part
j
to PU
i
(AGV ID k=1 for AGV
number 1, k=2 for AGV number 2 and so on.). Available parts in
the buffer area of any PU can be defined by the term C
i,j
= available
part
j
at any time in the buffer area of the PU
i
.
Figure 1: Shopfloor facility layout
2.2 Real-time Scheduling Problem of AGV
The primary challenge we are tackling is to ensure that parts are
delivered to each production unit (PU) on time. Specifically, we
aim to deliver each part
j
via an AGV to any PU
i
just before C
i,j
=0,
in order to avoid PUs idling without the necessary parts, while
also preventing AGVs from waiting until the buffer area is empty
494