The paper provides a complete procedure for simulating the behavior of a user-designed fluid CPS system in the Cyber Physical Systems Laboratory-as-a-Service (CPS-LaaS) framework. The operation of the system is based on the logical layout of the transport system.
Solution Summary
So the user's transport system designs are simply sent to the cloud and their behavior is simulated by those typical robots. And finally, the performance of the systems is returned to the user's screen from web-based lab backend.
Cyber-Physical Systems
Model-Integrated Computing
In MIC, we use meta-languages to represent key components of system information through modeling. The reason GME is flexible is that it can use a meta-language to describe the concepts, relationships, and constraints of specific domains.
Graph Rewriting and Transformation
The mapping relationship of the input and output model is defined by the Attribute Mapping object. According to the content and type of message, the typical command can be transferred to each tank to simulate the actions of the conveyor system.
Basic Sample Model to Simulate
Overview of the Solution Architecture
According to Figure III.2, in the modeling part, there are two domains, the complex domain is used for the user to design their own design, and their design can be transformed into a shape defined in the simple domain, the Global Network. Therefore, we design a global network; this is a panel type with only one node type and needed.
Reconfigurable Conveyor System Design
Complex-domain Metamodel
Grid can generate code for Robocode for background simulation that the user will not know what is happening in the background, what they will get is the final java animation to see their system working. The system contains three different types of models, input, output and block, these three models are the key component to build a transport system. A packet is contained in an input, which is used to represent the different types of packets to be transferred in the running system.
Input and Output models are used intuitively to design the input ports and output port of a conveyor system; model Block is inherited by three child model SegmentWE, SegmentNS and Turnaround, indicating the first two of three children that the conveyor belt moves from west-east direction, the belt that moves from north-south direction. Turnover is a transit point, which acts as the connection between one tape and another tape segment to maintain or change the package's original direction as we discussed in the basic sample model. In the higher hierarchy (please see the red region in Figure III.4), System is referred to as the reference test system and embodied in the Experiment.
According to this connectivity, we can activate any number of input ports and output ports to transfer and receive the packet. It will generate significantly more permutation and combination of experiments if we use only one.
Complex Domain-specific Model
In the view of Experiment Model (Figure III.6), namely New Experiment, New System is presented as a reference in the middle, each input port is connected to one Input Model, and this is the only way to represent that the New System s input port can be activated. If there is no connection within the input ports and external models, the input ports of NewSystem will still be mapped to the ongoing process, but the node to mapping is no longer activated, namely it is invalid to use in simulation part, it will be in the description of the GREAT transformation is discussed. In this case all input ports are enabled and will be mapped, meanwhile all output ports are all enabled with the same meaning.
Eventually, during the transformation of this carrier system design into the global network, all enabled gateways, loops, and connections will be defined and valid.
Global Grid Design
Simple-domain Metamodel
In this way, the behavior of the connection can be easily observed in the network and it can be easily changed to suit other needs of another CPS domain. Since the node can play a role in the reversal, we set 4 models as the node door, the model east, west, north and south. East can only be connected to West and North can only be connected to South, in this case it ensures the justice of the grid direction and makes the whole grid as a vertical and horizontal plate.
The global grid instance is quite understandable, according to the DSML model of complex domain, 3 input ports, 3 output ports and 6 turnarounds will be depicted as the node of the grid, the segment bands are mapped as the connection within each node. The result is presented in Figure III.8 In the next part, we will present the details of how this mapping transformation works.
Graph Rewriting and Transformation
In the next step, we will map the conveyor belt design experiment onto a giant grid plate; we will discuss this in the future work section. Since the definition of a GReAT transformation is based on a metalanguage to create a connection at the base level between two models, we do not need a DSML model, since it is one instance of the metamodel, so it is not representative. However, the DSML model is used in a real transformation, and the ultimate goal is to produce a transformation on the DSML model.
Crosslink defines the inner association of an Association Relationship type connection to link the source metamodel component to the destination so that the components on the two sides of the link can fully bind to each other. For example, after reading a sample DSML transporter model and an Input model, create a new Node in the grid, since we have defined the cross-link (please see figure cross-link in Figure III.10), then the whole operation on the read input port will also affect and only affect the newly built Node. NewBlock contains the transformation logic; as follows Table III.1 presents the source components and destination components of the four main creation logic strategies.
After the four steps above, the DSML model in Figure III.5 will be transformed into the model shown in Figure III.8.
Robocode
Tank Design Description
For example, the location of the tank according to the coordination of the component in grid, the type of tank according to the logic from grid, the combat time must be simulated. Finally, we generate and output our destination file -- Robocode's '.battle' file, which records the tanks' type, initial location (the tank's position on the battlefield), initial heading (the heading determines the firing direction), and initial bearing. radar direction (the direction of the sensor area. In this paper, we set the radar direction is the same with heading). Meanwhile, all the other tanks are set as a team, and within the team, each tank can communicate with each other using message sending and receiving.
It can only receive a message from the conveyor belt New Tank if the tank is facing the turnaround and send its message to the conveyor tank it is facing. The direction of the tank intersection is along horizontal or vertical lines facing the inlet/outlet openings or other intersection. The internal logic and total number of input and output ports of the current system design select the type of group leader.
The team leader is chosen according to the number of input ports and output ports.
Safety Path
- Create the safety path
- Cyber & Physical Simulation Feature
- Bullet shooting
First of all, the tank team leader acquires the map of the following packet type for each input ports. The tank input port is indexed by a natural number, so the lower the number of input ports, the higher the transmission priority is given. Then, the tank junction that received the packet size message will reverse its course according to the value of the y-coordinate.
After the heading is decided, the tank crossing scans its nearest tank conveyor, and asks it to stop and wait next to the tank crossing itself. Repeat this process until the tank crossing has found the final destination tank outlet, after which a safety road is built. Then the tank conveyor will start moving back and forth to stop the conveyor moving.
Tank-Intersection The tank is hit by the tank-Conveyor from behind, representing that it receives the package, and shoots to the tank-Conveyor its turn to, which means that the package is sent to the next segment.
Java Animation double-buffering
GReAT Transformation Time
Simulation Limitations
- Tank Life
According to the general rule above, we require each tank's initial health energy to be 100. If a tank continuously hits another tank, then the shooting tank's energy will keep increasing by 2*bullet power. However, if the tank is constantly under attack, the total energy value will be reduced by 2*bullet + 2*ceiling *(bullet power - 1).
Correspondingly, the tank-carrying, tank-crossing and tank-output types of tanks will gradually lose their energy until the life energy is evacuated. So for the CPS students' design, they can't put in too many packets, because if the life energy of the tank is zero, then the tank will disappear, which will destroy the original logic path. Because of rules 2 and 3, after the tank fires, the gun barrel will generate heat to limit the tank's firing frequency.
Therefore, in a real-time distributed cyber-physical system, this delay can cause errors.
Design Strategy
Simulation Improvement
- Multiple packages
- Using a Physical Engine
- Refine the animation
- Failure Tolerance
The first plan of the approach is based on the assumption that the global network space is large enough to handle all the experiments. To calculate the efficiency of the scheduling methods, we first calculate the area utilization of the total network nodes in use. The second approach focuses on the space limitations of the global network, so in some parts of the network, nodes must be separated by multiple experiments, as Figure V.3 shows.
Due to the convenience of the Internet, we can handle a wide range of requirements from CPS students to simulate their system designs, for example a system design of a reconfigurable conveyor system. Our complex system design and simple system design are all instances of the corresponding domain metalanguage defined in GME, and the transformation is done by GReAT. The final animation as the execution of the student's experiment design will be presented on the student's screen, which is generated by.
Branicky, “CPS Laboratory-as-a-Service: Enabling Technology for Easily Accessible and Scalable CPS Education,” in Proceedings of the First Workshop on Cyber-Physical Systems Education (CPS-Ed) at CPSWeek 2013, Philadelphia, PA, USA :IEEE, Apr.
