This report consists of 47 pages and 4 appendices. It may only be reproduced literally and as a whole. For commercial purposes only with written authorization of Delft University of Technology. Requests for consult are only taken into consideration under the condition that the applicant denies all legal rights on liabilities concerning the contents of the advice.
Delft University of Technology FACULTY MECHANICAL, MARITIME AND MATERIALS ENGINEERING
Department Maritime and Transport Technology Mekelweg 2 2628 CD Delft the Netherlands Phone +31 (0)15-2782889 Fax +31 (0)15-2781397 www.mtt.tudelft.nl
Specialization: Transport Engineering and Logistics Report number: 2017.TEL.8176
Title: Safety Protection for automated guided vehicles
Author: Jesse Kwakkenbos
Assignment: Literature Assignment (ME54010) Confidential: No
Supervisor: dr.ir. Y. Pang Date: November 29, 2017
Delft University of Technology FACULTY MECHANICAL, MARITIME AND MATERIALS ENGINEERING
Department Maritime and Transport Technology Mekelweg 2 2628 CD Delft the Netherlands Phone +31 (0)15-2782889 Fax +31 (0)15-2781397 www.mtt.tudelft.nl
Student: Jesse Kwakkenbos Assignment type: Literature Supervisor: Dr.ir. Y. Pang Credit points (EC): 10
Specialization: TEL
Report number: 2017.TEL.8176
Confidential: No
Subject: Safety Protection for automated guided vehicles
Nowadays, there are many systems that make use of different applications of automated guided vehicles. There are multiple reasons to use automated guided vehicles in such systems, some benefits of the use of these vehicles: cost reduction, reduce human labor, increase capacity, accuracy, reliability, scalability and flexibility. Prior goal of these systems with operating automated guided vehicles is the enforcement of safety.
The assignment is to look for and understand the different safety systems that are used for a safe operation in an environment with automated guided vehicles. How are these systems working and communicating to maintain safety, which sensors are used for these automated guided vehicles and which algorithms or methods are used to transmit the data received from the sensors. It is also important to look to the implementation of the safety standards, a requirement in these industries
The report should comply with the guidelines of the section. Details can be found on the website. The mentor,
i
Summary
Automated guided vehicles are mostly used in industries, they are controlled by different algorithms and there is no driver on board. There are multiple reasons to use automated guided vehicles in a system; cost reduction, reduce human labor, increase capacity, accuracy, reliability, scalability and flexibility. To determine the position of automated guided vehicles are two systems used. A centralized system is used, where the vehicle receives information about its position from the control room. Decentralized systems are also used, where the vehicle determines its own position. The automated guided vehicle will use sensors to determine the position in this system.
There are different types of automated guided vehicles which are operating in different systems. Some different types: carriers, tuggers, forklifts, unit load carriers, people movers and automated lifting
vehicles. These types of vehicles are used for different applications. Automated guided vehicles are mostly used for transshipment, storage, transport or production.
A safe operating system with automated guided vehicles is desired, but for possible safety accidents are three different types of conflicts introduced:
• Dynamic environment: multiple automated guided vehicles are operating in this environment, conflicts could occur between automated guided vehicles and conflicts between automated guided vehicles and static obstacles
• Static environment: a single automated guided vehicle is operating in a system and conflicts could occur between the vehicle and static obstacles
• Conflicts between automated guided vehicles and human being
To take care of a safe operating system, the automated guided vehicles are equipped with a lot of sensors. These sensors can recognize static or dynamic objects that are in the surrounding of the operating automated guided vehicles, they are mostly equipped with:
• Ultrasonic sensors • Surface sensors • Safety edges sensors • Lasers
• Time-of-flight cameras • Bumper sensors • Emergency-stop button
ii
For a safe control of automated guided vehicles, distinction in three controls is made; collision avoidance control, zone control and combination control. For collision control are different methods used to maintain safety:
• Petri Net Method
• Game Theory combined with Modern control • Time Window Method
• Dijkstra
• Ultra-wide band
Zone control eases the avoidance of automated guided vehicle collisions by demanding that each zone can be occupied by at most one vehicle. Zone control is also important to resolve possible conflicts and accidents between automated guided vehicles in a system. Combination control is collision avoidance control combined with zone control.
To maintain safe operation areas, different standards are used for automated guided vehicles. Laws and regulations, safety standards and guidelines are used to take care about the safety.
iii
Figures & Tables
Figures
figure 1: Automated guided vehicle [3] ... 4
figure 2: WEpod Wageningen [6] ... 5
figure 3: Automated guided vehicle in production area [8] ... 6
figure 4: The schematic diagram of the intersection of conflict situations [9]... 7
figure 5: The schematic diagram of the catching up of conflict situations [9]... 7
figure 6: The schematic diagram of opposite conflict situations [9] ... 8
figure 7: Visual working of ultrasonic sensor [11] ... 11
figure 8: Triangulation [14] ... 15
figure 9: Point clouds generated for Radiohead's House of Cards music video, which used lasers instead of cameras [14] ... 15
figure 10: Different warning and protection fields [17] ... 16
figure 11: 3D time-of-flight camera operation [18] ... 17
figure 12: Two time-of-flight methods: pulsed (top) and continuous-wave (bottom) [18] ... 18
figure 13: Extending distance using a multi-frequency technique [18] ... 20
figure 14: Depth map of soda cans [18] ... 20
figure 15: Avatar formed from point-cloud [18] ... 21
figure 16: Safety Bumpers [12] ... 21
figure 17: Example of a safety edge sensor [21] ... 22
figure 18: Working of safety edge [21] ... 23
figure 19: Emergency stop button [23] ... 23
figure 20: Scheme of Petri net [24] ... 24
figure 21: The schematic diagram of collision avoidance simulation situations [9]... 27
figure 22: The schematic diagram of opposite conflict situations [9] ... 28
figure 23: The schematic diagram of opposite of conflict time window [9] ... 29
figure 24: The schematic diagram of the catching up of conflict situations [9] ... 29
figure 25: The schematic diagram of the catching up of conflict time window [9] ... 29
figure 26: Shortest path calculation according Dijkstra’s algorithm [28] ... 30
figure 27: Representation of a 2D proposed warehouse [29] ... 32
figure 28: Flowchart representation of the routing task [29] ... 33
figure 29: Multiple stuck automated guided vehicles [31] ... 35
figure 30: Advanced perception system [31] ... 36
figure 31: Grid with target nodes [1] ... 38
iv
Tables
v
Content
Summary ... i
Figures & Tables ... iii
Figures ... iii
Tables ... iv
Content ... v
1. Introduction ... 1
2. Automated guided vehicles ... 2
2.1. Working of automated guided vehicle ... 2
2.2. Different applications automated guided vehicles ... 3
2.2.1. Transshipment ... 3
2.2.2. Storage ... 4
2.2.3. Transport ... 4
2.2.4. Production ... 5
2.3. Different safety accidents ... 6
2.3.1. Collision between automated guided vehicles ... 6
2.3.2. Collision between automated guided vehicles and static object... 8
2.3.3. Collision between automated guided vehicle and human being ... 8
3. Safety components ... 11
3.1. Non-contact distance sensors ... 11
3.1.1. Ultrasonic sensors ... 11
3.1.2. Surface pressure sensors ... 12
3.1.3. Lasers ... 13
3.1.4. Time-of-Flight cameras ... 17
3.2. Bumper sensors ... 21
3.2.1. Safety bumpers ... 21
3.2.2. Safety edges sensors ... 22
vi
4. Safety control ... 24
4.1. Collision avoidance control ... 24
4.1.1. Petri net method ... 24
4.1.2. Game theory combined with modern control theory... 26
4.1.3. Time window method ... 28
4.1.4. Dijkstra’s Algorithm ... 30
4.1.5. Collision avoidance using hulls ... 33
4.1.6. Ultra-wide band technology ... 37
4.2. Zone-control ... 39
4.3. Combination control ... 43
5. Safety standards ... 44
5.1. Laws and Regulations ... 44
5.2. Standards ... 44
5.3. Guidelines ... 45
6. Conclusion ... 46
References ... 48
Appendix A: DIN EN1525 ... 52
Appendix B: DIN EN ISO 3691 ... 53
Appendix C: DIN EN 1175-1 ... 55
Introduction
2017.TEL.8176 1
1. Introduction
This literature assignment is about the safety protection for automated guided vehicles. The abbreviation of an automated guided vehicle is AGV, the full term appears in the main content of the report and afterwards only the abbreviation is used.
The different kinds of AGVs are discussed, the way the AGVs are operating in different businesses. In this report is talked about the different kinds of AGVs and the different applications where the AGVs are used for. The meaning of safety will also be discussed, for example, a collision between AGVs is different than a collision between an AGV and a human. What sensors are used for safety systems are still on the market and how are these systems safely controlled. It is also important to follow the different safety standards that are introduced for AGVs. The use of safety standards will improve the safety in environments with operating AGVs.
The goal of this assignment is to achieve a better understanding of the working of the safety protection for AGVs and the different systems that are used for different operating networks with AGVs. Two parts of this report are the most important subjects about the safety protection for AGVs
• Safety components • Safety control
These two subjects are mentioned after a general chapter about AGVs. In this general chapter is the working of an AGV explained. The different applications where AGVs are used for and the different safety accidents that can occur are also discussed in this chapter. In safety components are the different sensors and bumpers and the working of these parts explained. The chapter safety control will explain how a safe system with operating AGVs is obtained and how the information retrieved from the sensors is used. As last the different safety standards that are used for operating systems with AGVs are discussed.
Automated guided vehicles
2017.TEL.8176 2
2. Automated guided vehicles
AGVs are vehicles that are used in industries mostly. The AGVs are controlled by different algorithms, there is no driver on board.
The next properties and functions of AGVs are retrieved from Mark Duinkerken [1].
AGVs are introduced for the first time in 1953. This first AGV was used to pull a trailer and follow an overhead wire in a grocery warehouse. In 1973 the first assembly vehicle was introduced by Volvo in Sweden. Nowadays multiple AGVs are operating automatically in independent systems. There are multiple reasons to use AGVs in a system, these reasons are mentioned below:
• Cost reduction • Reduce human work
o Shortage of human labor o Unsafe environments o Prevent routine work • Increase capacity
• Accuracy • Reliability
• Scalability and flexibility
The safety aspect of the AGVs is important in this literature assignment. Safety is split up in three categories, first the safety for humans working in AGV systems, second the AGVs self in the system and last the AGVs compared to static objects. These categories are handled in chapter 2.3.
2.1. Working of automated guided vehicle
To describe the working of the AGV are different hardware components used, the components are retrieved from Mark Duinkerken [1].
There are different AGV types, for example: standard carriers, tuggers, forklifts, unit load carriers, people movers and ALVs (Automated Lifting Vehicles). All these types have different driving characteristics and driving technologies. The different applications where these types are operating are discussed in the next section (section 2.2.).
The safety for AGVs is also important, the safety is managed by external control. This control uses different safety sensors to maintain safety, these sensors are mentioned in section 3. The human factor plays also an important role in the safety for the AGVs. Different kind of accidents are mentioned in section 2.3.
Automated guided vehicles
2017.TEL.8176 3
For the infrastructure are topology and docks necessary. The topology exists of a map with boundaries, permanent obstacles, one/two-way road, road crossings and locations of the docks. Distinction is made for different kinds of docks, pick-up and delivery points, charging/fueling locations, maintenance locations and locations for the idle vehicles.
Even the communication is necessary in a system with AGVs, for the communications between AGVs and the infrastructure are different components used, some of the different communication types; detections wires, lights signals, FM radio signals, wireless LAN, Bluetooth.
For the navigation of AGVs is a guide path needed, inertial guidance is also used to determine the position. Closed loop control is used for position updates, for the closed loop control is a centralized positioning system or an autonomous position determination used.
For the determination of the position of the AGV are the following systems used [2]: • Centralized (AGV receives position information)
o Natural feature: Reference images are recorded and stored, the position of the AGV is calculated based on its relative position compared to those natural features. • Decentralized (AGV determines own position)
o Laser triangulation: A laser scanner is scanning for reflective targets that are mounted on fixed positions, the vehicle control algorithms can calculate the exact position.
o Magnetic: Magnetic tape is applied to the surface of the floor, a sensor can detect the magnetic tape.
o Wire: A wire embedded in the floor is used to determine the location. Antennas detect the signal from the wire and the encodes can calculate the distance. With multiple signal the position can be determined.
o Optical: Chemical or tape strip is applied to the floor, the AGV has sensors which can detect the path.
For the detection of obstacles, AGVs mostly equipped with lasers and mechanical bumpers.
2.2. Different applications automated guided vehicles
AGVs are used if different applications areas, the main applications are discussed below
2.2.1. Transshipment
AGVs that are used for transshipment are mostly operating in terminals. The AGVs used in port container terminals are fully automated for handling 20’, 40’ and 45’ ISO containers. The AGV can handle loads of
Automated guided vehicles
2017.TEL.8176 4
up to 70 tons with a maximum speed of seven meters per second. Expertise leads to good performing container carrier AGV with good characteristics regarding fuel efficiency, CO2 emissions, noise levels, reliability and maintainability. The result of these efforts is an environmentally friendly AGV with low operating costs and a high MTBF rate, mean time between failure rate. The next sources are used for the information above: VDL Group [3] and Konecranes [4].
figure 1: Automated guided vehicle [3]
2.2.2. Storage
The source Dematic Egemin automation [5] declares that AGVs are mostly used in warehouses and distribution centers for a lot of storage applications, they are used to reduce labor costs and the efficiency and reliability of the storage process is increased in most cases. AGVs can be used for very narrow aisles, but also for high storing loads (6-12 meter). A few examples of storage applications with AGVs are mentioned below:
• Block storage
• Pallet storage in warehouse racks • Floor-based deep stack storage of pallets • Vertical storage of reels
• Horizontal storage of reels in cradles
For a good working of these applications are some components used; the navigation sensor, a system used for running the sensor, safety scanners, pallet position sensors to find the position of items precisely. Most of these components are explained later.
2.2.3. Transport
AGVs are also used for transport, for example the transport of people (figure 2). The AGV will take care of a safe and reliable transportation along designated routes. AGVs are also used as transport equipment for
Automated guided vehicles
2017.TEL.8176 5
the distribution of goods [1].
figure 2: WEpod Wageningen [6]
2.2.4. Production
AGVs are even used for production applications [7].
A production process consists of mobile robots, better known as AGVs, used for transportation and automatic material handling; for example, finished goods, raw materials and products in process. The AGV is a driverless vehicle that performs the tasks of handling of flexible materials and is therefore considered suitable for an FMS (Flexible Manufacturing Systems) environment.
The AGV has the function to ensure efficient flow of materials within the production system. Production systems must be flexible and must allow the dynamic reconfiguration of the system. The AGV is a key component to achieve the objectives of an FMS. This means that the AGV should provide the required materials to the appropriate workstation, at the right time and in the right amount. Otherwise the production system will not perform well, making it less efficient, generating less profit or increasing the operating costs.
Automated guided vehicles
2017.TEL.8176 6
figure 3: Automated guided vehicle in production area [8]
2.3. Different safety accidents
According to Juntao Li et al. [9]. AGV types of conflict can be divided into three categories in different types of environment:
• In a dynamic environment, the quantity of AGV is greater than one, there are some conflicts between AGV and the others and the conflicts which between AGV and the static obstacles. • In a static environment and the quantity of AGV is one, the conflict which needs to be resolved is
the collision between AGV and static obstacles
• A conflict between an AGV or multiple AGVs with human being
2.3.1. Collision between automated guided vehicles
In a dynamic environment, conflicts between AGV and the others can be divided into three basic types: • Including the intersection of the conflict
• The catching up of the conflict • The opposing of the conflict
The three different types will be discussed below, these types were retrieved from Juntao Li et al. [9]. The intersection of conflict
The intersection of conflict does not belong to the catching up of conflict and the opposing of conflict. The two AGVs which create the conflict are neither positive relative nor in the same lane, that is crossing collision. The scenario of conflict is shown in the following figure:
Automated guided vehicles
2017.TEL.8176 7
figure 4: The schematic diagram of the intersection of conflict situations [9] The catching up of conflict
The catching up of conflict here is not the conflict in the conventional sense, because the AGV is set to a constant speed in this study, the issue that AGV2 is slower than AGV1 in speed does not exist. In this case, when AGV2 needs to in-situ steer, it will waste more time than pass through the node straightly, then the conflict may occur. The scenario of conflict is shown in the following figure:
figure 5: The schematic diagram of the catching up of conflict situations [9] The opposing of conflict
The opposing conflict refers to the case that multi-vehicle AGVs meet head-on in the process of running. AGV1 and AGV2 are in opposite direction of movement and in the same lane. The scenario of conflict is shown in the following figure:
Automated guided vehicles
2017.TEL.8176 8
figure 6: The schematic diagram of opposite conflict situations [9]
If there is no corresponding avoidance strategy, whether AGVs are in a dynamic or a static environment, they will collide. In the multi-AGVs dynamic environment, based on the differences in different
environments and task distribution, there will be a variety of complex and diverse conflicts. But these conflicts are composed of the three most basic conflicts above. The essence of these three types of conflicts is the problem of temporal and spatial overlapping. It can be understood as, two AGVs occupy the same node at the same time, thus they collide. So, to resolve the complex conflicts in the case of multi-vehicle AGVs, ultimately these three types of conflicts should be resolved, and then to ensure the AGVs can move safely and smoothly [9].
2.3.2. Collision between automated guided vehicles and static object
AGVs are mostly equipped with sensors and are navigated by a path (section 2.1.). If the sensors are working properly, the probability of a collision with a static object is minimized. The path is installed before the start of use of AGVs and the whole operation area have a certain layout. If the AGV is following the path, the probability to have a collision with a static object that belongs to the operation area is minimized. When a static object, that does not belong the operation area, is left on the path, for example a pick-up trolley, it is important that the sensors of the AGV detect this object. The working of this principle is explained further in section 3.1.
2.3.3. Collision between automated guided vehicle and human being
The prevent collisions between AGVs and human beings, the AGVs are using sensors to detect the human being if they are in the same area as the AGV. The AGVs will use different warning zones to prevent such collisions, these zones will be explained later (section 3.1.3.)
AGVs are safer than vehicles which are controlled by humans. But sometimes accidents occur even for AGVs, an example of an accident with an AGV and a human being is stated below. This example of an accident is retrieved from [10].
Automated guided vehicles
2017.TEL.8176 9
It happened at a Kraft Food warehouse in Granite City, Ill. Early Tuesday morning an employee was discovered pinned between a laser-guided AGV and a metal racking unit. This employee was responsible for overseeing the operation of the facility's AGVs. OSHA and the Granite City Police Department are still investigating this accident, but one thing seems certain: if proper safety procedures were followed this shouldn't have happened. There's a standard that vehicle manufacturers and users are supposed to follow: The ANSI/ITSDF B56.5-2012 Safety Standard for Driverless, Automatic Guided Industrial Vehicles and Automated Functions of Manned Industrial Vehicles (see section 0.).
Some statements from the research:
‘Automatic guided industrial vehicles can cause injury or damage if improperly used or maintained and if the potential risks specified in user training associated with hazard zones and restricted areas are not respected by persons within or adjacent to these areas.’
It also states what's supposed to happen with a properly equipped AGV.
‘The braking system in conjunction with the object detection system and the response time of the safety control system shall cause the vehicle to stop prior to impact between the vehicle structure and other mounted equipment, including its intended load, and an obstruction being sensed in advance of the moving vehicle in the main direction of travel.’
‘Although the vehicle braking system may be performing correctly and as designed, it cannot be expected to function as designed and specified should an object suddenly appear in the path of the vehicle and within the designed safe stopping distance. Examples include, but are not limited to, an object falling from overhead or a pedestrian stepping into the path of a vehicle at the last instant.’
• Concluded from this research: training is so critical for both operators and the people working around these vehicles. This can be reached by:
• A training program for operators and other user personnel likely to be exposed to the system in operation, including visitors, shall include the system supplier's documented operating instructions and procedures and the user's local applicable requirements if any.
• The initial training shall be presented by the system supplier to all operators and other user personnel and not condensed or eliminated for those claiming previous experience.
• Oral, written, or operational performance tests and evaluations should be given during and at the completion of all training.
Automated guided vehicles
2017.TEL.8176 10
But it's the failure to heed the following pointer that explains why fatalities are still associated with industrial trucks:
• Periodic, ongoing training or refresher training sessions shall then be conducted by the user for the benefit of existing users as well as for new user personnel and visitors. Refresher training sessions, which may be condensed versions of the initial training sessions, and periodic on-the-job evaluation, are as important as initial training, especially when new personnel are hired or otherwise introduced to the system following initial deployment.
The ANSI standard defines fail-safe as a design in which no single failure can cause an unsafe condition. It is eventually known whether the victim in this case might have been involved in multiple failures at one time, but this should be a wake-up call for every user of automated material handling equipment: Never fail to train.
Safety components
2017.TEL.8176 11
3. Safety components
3.1. Non-contact distance sensors
Different non-contact distance sensors are used to maintain safety for AGVs, the most used non-contact sensors are mentioned below:
3.1.1. Ultrasonic sensors
The definition of an ultrasonic sensor is retrieved from the website Carnegie Mellon Robotics Academy [11].
An ultrasonic sensor is a device that can measure the distance to an object by using sound waves. It measures distance by sending out a sound wave at a specific frequency and listening for that sound wave to bounce back. By recording the elapsed time between the sound wave being generated and the sound wave bouncing back, it is possible to calculate the distance between the sonar sensor and the object.
figure 7: Visual working of ultrasonic sensor [11] 1. 𝐷𝑖𝑠𝑡𝑎𝑛𝑐𝑒 = 𝑠𝑝𝑒𝑒𝑑 𝑜𝑓 𝑠𝑜𝑢𝑛𝑑∗𝑡𝑖𝑚𝑒
2
The speed of sound through air is 344 m/s. If the time of the returning wave is measured and this time is multiplied by the speed of sound, the round-trip distance of the sound wave is known. Round-trip means that the sound wave traveled two times the distance to the object before it was detected by the sensor; it includes the 'trip' from the sonar sensor to the object and the 'trip' from the object to the ultrasonic sensor. To find the distance to the object, simply divide the round-trip distance in half.
It is important to understand that some objects might not be detected by ultrasonic sensors. This is because some objects are shaped or positioned in such a way that the sound wave bounces off the object, but are deflected away from the ultrasonic sensor. It is also possible that the object is too small to be detected, because there is not enough reflection. Other objects can absorb the sound wave all
Safety components
2017.TEL.8176 12
together, which means that there is no way for the sensor to detect them accurately. These are important factors to consider when designing and programming a robot using an ultrasonic sensor.
According to Mayser Industry & Mechanical Engineering [12] is ultrasonic monitoring of surroundings, entry areas and other areas the ideal solution for non-contact detection of people and objects, as well as for distance measurement. If a person or object is detected in the area monitored, the system can slow down or stop an automatic movement. Even the smallest objects are reliably detected over the entire distance, irrespective of material, shape, transparency or color. There are two important advantages of the use of ultrasonic sensors; first, the sensors are unaffected by dirt, noise, airflow and moisture. Operating without blind zone up to 2.5 meter is the second advantage.
3.1.2. Surface pressure sensors
According to Cynergy [13] measures a pressure sensor the pressure, typically of gases or liquids. Pressure is an expression of the force required to stop a gas or fluid from expanding, and is usually stated in terms of force per unit area. A pressure sensor generates a signal related to the pressure imposed. Typically, such a signal is electrical, but it might also include additional means, such as optic signals, visual signals and/or auditory signals.
Pressure sensors are used in numerous ways for control and monitoring in thousands of everyday applications. Pressure sensors can be used in systems to measure other variables such as fluid/gas flow, speed, water level, and altitude. Pressure sensors can alternatively called pressure transducers, pressure transmitters, pressure senders, pressure indicators among other names. Pressure sensors can vary drastically in technology, design, performance, application suitability and cost. A conservative estimate would be that there may be over 50 technologies and at least 300 companies making pressure sensors worldwide.
There is also a category of pressure sensors that are designed to measure in a dynamic mode for capturing very high-speed changes in pressure. Example applications for this type of sensor would be in the measuring of combustion pressure in an engine cylinder or in a gas turbine. These sensors are commonly manufactured out of piezoelectric materials like quartz. Some pressure sensors function in a binary manner, for example, when pressure is applied to a pressure sensor, the sensor acts to complete or break an electrical circuit. Some speed cameras use this binary manner. These types of sensors are also known as a pressure switches.
Pressure sensors can be classified in term of pressure ranges they measure, temperature ranges of operation, and most importantly the type of pressure they measure. In terms of pressure type, they are categorized in five categories [13]:
Safety components
2017.TEL.8176 13
• Absolute pressure sensor: This sensor measures the pressure relative to perfect Vacuum pressure (0 PSI or no pressure). Atmospheric pressure, is about 100kPa (14.7 PSI) at sea level. Atmospheric pressure is an absolute pressure.
• Gauge pressure sensor: This sensor is used in different applications because it can be calibrated to measure the pressure relative to a given atmospheric pressure at a given location. An example of gauge pressure would be a tire pressure gauge. When the tire pressure gauge reads 0 PSI, there is really 14.7 PSI (atmospheric pressure) in the tire.
• Vacuum pressure sensor: This sensor is used to measure pressure less than the atmospheric pressure at a given location.
• Differential pressure sensor: This sensor measures the difference between two or more pressures introduced as inputs to the sensing unit. For example, measuring the pressure drop across an oil filter. Differential pressure is also used to measure flow or level in pressurized vessels.
• Sealed pressure sensor: This sensor is the same as the Gauge pressure sensor except that it is previously calibrated by manufacturers to measure pressure relative to sea level pressure (14.6 PSI).
In combination with operating AGVs, pressure sensitive surface sensors detect presence in dangerous areas of movement, for instance on machines or in spaces used in collaboration in different applications. The presence of a person or object in the protection zone slows down or stops the movement of the AGV. Maintenance-free, robust system setup, resistant to external influences and reliable functioning in dirty environment conditions are benefits of surface sensors [12].
3.1.3. Lasers
The working principle of lasers is described by Ian Wright [14]. At the most basic level, laser scanning involves a combination of controlled steering of laser beams and distance measurements. A rotary encoder controls the scanning motion by adjusting multiple scanning mirrors to guide the beams. Positioning the laser beam in two dimensions requires either rotating a single mirror along two axes or, for faster scanning, reflecting the beam onto two closely spaced mirrors mounted on orthogonal axes. The lasers can also be positioned in three dimensions via a servo-controlled lens system known as a focus- or Z-shifter.
Single scans are generally not sufficient to produce a complete model of the subject and so multiple scans must be brought together into a common reference system through a process known as alignment or registration.
Safety components
2017.TEL.8176 14
Time-of-Flight
One of the central concepts in laser scanning is time-of-flight. This refers to the use of a laser range-finder to time the round trip of a light pulse travelling from the scanner to the object and back again. Since the speed of light is a known constant, the distance between the scanner and the object can be calculated by measuring the time it takes for the light pulse to return to the scanner. Thus, the formula for calculating distance based on time-of-flight is simply:
2. 𝑑 =𝑐∗𝑡 2 𝑑 = 𝑑𝑖𝑠𝑡𝑎𝑛𝑐𝑒 [𝑚] 𝑐 = 𝑠𝑝𝑒𝑒𝑑 𝑜𝑓 𝑙𝑖𝑔ℎ𝑡 [𝑚 𝑠] 𝑡 = 𝑡𝑖𝑚𝑒 𝑓𝑜𝑟 𝑡ℎ𝑒 𝑙𝑖𝑔𝑡ℎ 𝑡𝑜 𝑟𝑒𝑡𝑢𝑟𝑛 𝑡𝑜 𝑡ℎ𝑒 𝑠𝑐𝑎𝑛𝑛𝑒𝑟 [𝑠]
Therefore, the accuracy of a time-of-flight sensor depends on the accuracy of the laser scanner’s chronometer. The main advantage of this technique is its long-distance capability, which is why some time-of-flight scanners are used for large-scale surveying projects. However, accurately measuring a quantity as miniscule as the time it takes for a pulse of light to make a round trip in a quality lab is difficult. That is why time-of-flight sensors alone are insufficient for applications requiring high-accuracy measurements.
Triangulation
Many handheld scanners use triangulation to off-set the diminished accuracy that comes with time-of-flight measurements.
Safety components
2017.TEL.8176 15
figure 8: Triangulation [14]
For example, handheld laser scanners often incorporate a camera that tracks the laser dots projected onto scanned subjects. This enables the scanners to triangulate the distance of the subjects much more accurately than with time-of-flight measurements alone. However, although triangulation enables higher accuracy, it also limits the effective range of the scanner.
Point Clouds and Registration
In the most basic terms, a point cloud refers to a set of data points in a coordinate system. In the standard Cartesian coordinate system, points are defined in terms of X, Y and Z coordinates. In the context of 3D scanning, point clouds represent the results of a scan as unstructured three-dimensional data. The typical file formats of point clouds are TXT, IGS and ASCII.
figure 9: Point clouds generated for Radiohead's House of Cards music video, which used lasers instead of cameras [14]
Point cloud data is then brought into a common reference system where the data is merged into a complete model via the process of alignment or registration. This process can occur during the scan itself (in the case of high-end 3D scanners) or as a post-processing step. The resulting data can be processed further using software to clean up aberrations or fill in gaps in the data points.
Safety components
2017.TEL.8176 16
Lasers in combination with AGV
Laser scanners from SICK Insight [15] and RS [16] are used to provide collision protection from AGVs (AGVs). Lasers scanners can enable safe travelling along complex paths and production layouts with safe non-contact detection of personnel. When there are certain lasers used, a scanning angle of 360 degrees is created. All-round monitoring of vehicles is now created. For the all-round monitoring are sixteen flexibly configurable triple field sets used, each field has two warning and one protective zone, the three different zones are described below;
• Outer zone: If a pedestrian enters, a warning alarm may be sounded • Middle zone: The vehicle may decelerate
• Inner zone: An emergency stop can be triggered if this zone is entered.
The following example of the working of the laser scanner is retrieved from another manufacturer, Jungheinrich [17].
If a person enters the warning field (outer zone), an alarm is used. When this person remains in this field, the AGV immediately slows down (middle zone). If the person comes to close to the AGV, the AGV will stop (inner zone), see figure 10. With these zones is the need to stop the vehicle for every object detection minimized, even as the affection of the production efficiency. This system has an increased detection of the safety field up to seven meters by continuing development.
Safety components
2017.TEL.8176 17
3.1.4. Time-of-Flight cameras
The next information about time-of-flight cameras is retrieved from a paper from Larry Li [18].
3D Time-of-Flight (TOF) technology is revolutionizing the machine vision industry by providing 3D imaging using a low-cost CMOS pixel array together with an active modulated light source. Compact construction, easy-of-use, together with high accuracy and frame-rate makes TOF cameras an attractive solution for a wide range of applications. The TOF operation will be covered, and compare TOF with other 2D/3D vision technologies. Then various applications that benefit from TOF sensing, such as gesturing and 3D scanning and printing, are explored. Finally, resources that help readers get started with Texas Instruments’ 3D TOF solution are provided.
Theory of Operation
A 3D time-of-flight (TOF) camera works by illuminating the scene with a modulated light source, and observing the reflected light. The phase shift between the illumination and the reflection is measured and translated to distance. The basic TOF concept is illustrated in figure 11. Typically, the illumination is from a solid-state laser or a LED operating in the near-infrared range (around 850 nm) invisible to the human eyes. An imaging sensor designed to respond to the same spectrum receives the light and converts the photonic energy to electrical current. Note that the light entering the sensor has an ambient component and a reflected component. Distance (depth) information is only embedded in the reflected component. Therefore, high ambient component reduces the signal to noise ratio (SNR).
figure 11: 3D time-of-flight camera operation [18]
To detect phase shifts between the illumination and the reflection, the light source is pulsed or modulated by a continuous-wave (CW), source, typically a sinusoid or square wave. Square wave modulation is more common because it can be easily realized using digital circuits.
Safety components
2017.TEL.8176 18
Pulsed modulation can be achieved by integrating photoelectrons from the reflected light, or by starting a fast counter at the first detection of the reflection. The latter requires a fast photo-detector, usually a single-photon avalanche diode (SPAD). This counting approach necessitates fast electronics, since achieving 1 millimeter accuracy requires timing a pulse of 6.6 picoseconds in duration. This level of accuracy is nearly impossible to achieve in silicon at room temperature.
figure 12: Two time-of-flight methods: pulsed (top) and continuous-wave (bottom) [18] The pulsed method is straightforward. The light source illuminates for a brief period (∆t), and the reflected energy is sampled at every pixel, in parallel, using two out-of-phase windows, C1 and C2, with the same ∆t. Electrical charges accumulated during these samples, Q1 and Q2, are measured and used to compute distance using the formula:
3. 𝑑 =1 2𝑐∆𝑡 (
𝑄2 𝑄1+𝑄2)
In contrast, the CW method takes multiple samples per measurement, with each sample phase-stepped by 90 degrees, for a total of four samples. Using this technique, the phase angle between illumination and reflection, φ, and the distance, d, can be calculated by
4. 𝜑 = arctan (𝑄3−𝑄4 𝑄1−𝑄2)
Safety components
2017.TEL.8176 19
5. 𝑑 = 𝑐 4𝜋𝑓𝜑
It follows that the measured pixel intensity (A) and offset (B) can be computed by: 6. 𝐴 =√(𝑄1−𝑄2)2+(𝑄3+𝑄4)2
2
7. 𝐵 =(𝑄1+𝑄2+𝑄3+𝑄4)
4
In all the equations, c is the speed-of-light constant.
At first glance, the complexity of the CW method, as compared to the pulsed method, may seemed unjustified, but a closer look at the CW equations reveals that the terms, (Q3 – Q4) and (Q1 – Q2) reduces the effect of constant offset from the measurements. Furthermore, the quotient in the phase equation reduces the effects of constant gains from the distance measurements, such as system amplification and attenuation, or the reflected intensity. These are desirable properties.
The reflected amplitude (A) and offset (B) do have an impact the depth measurement accuracy. The depth measurement variance can be approximated by:
8. 𝜎 = 𝑐 4√2𝜋𝑓∗
√𝐴+𝐵 𝑐𝑑𝐴
The modulation contrast, cd describes how well the TOF sensor separates and collects the photoelectrons. The reflected amplitude, 𝐴, is a function of the optical power. The offset, 𝐵, is a function of the ambient light and residual system offset. One may infer from equation 8 that high amplitude, high modulation frequency and high modulation contrast will increase accuracy; while high offset can lead to saturation and reduce accuracy. At high frequency, the modulation contrast can begin to attenuate due to the physical property of the silicon. This puts a practical upper limit on the modulation frequency. TOF sensors with high roll off frequency generally can deliver higher accuracy.
The fact that the CW measurement is based on phase, which wraps around every 2π, means the distance will also have an aliasing distance. The distance where aliasing occurs is called the ambiguity distance, damb and is defined as:
9. 𝑑𝑎𝑚𝑏 = 𝑐 2𝑓
Since the distance wraps, damb is also the maximum measurable distance. If one wishes to extend the measurable distance, one may reduce the modulation frequency, but at the cost of reduced accuracy, as according to equation 9.
Safety components
2017.TEL.8176 20
Instead of accepting this compromise, advanced TOF systems deploy multi-frequency techniques to extend the distance without reducing the modulation frequency. Multi-frequency techniques work by adding one or more modulation frequencies to the mix. Each modulation frequency will have a different ambiguity distance, but true location is the one where the different frequencies agree. The frequency of when the two modulations agree, called the beat frequency, is usually lower, and corresponds to a much longer ambiguity distance. The dual-frequency concept is illustrated below.
figure 13: Extending distance using a multi-frequency technique [18]
In TOF sensors, distance is measured for every pixel in a 2D addressable array, resulting in a depth map. A depth map is a collection of 3D points (each point also known as a voxel). As an example, a QVGA sensor will have a depth map of 320 x 240 voxels. 2D representation of a depth map is a gray-scale image, as is illustrated by the soda cans example in figure 14, the brighter the intensity, the closer the voxel. The depth map of a group of soda cans is shown in figure 14.
figure 14: Depth map of soda cans [18]
Alternatively, a depth map can be rendered in a three-dimensional space as a collection of points, or point-cloud. The 3D points can be mathematically connected to form a mesh onto which a texture surface can be mapped. If the texture is from a real-time color image of the same subject, a life-like 3D rendering of the subject will emerge, as is illustrated by the avatar in figure 15. One may be able to rotate the avatar to view different perspectives.
Safety components
2017.TEL.8176 21
figure 15: Avatar formed from point-cloud [18]
3.2. Bumper sensors
3.2.1. Safety bumpers
Safety bumpers are an important safety component for protection against risk in a system with AGVs according to ISS Safety [19] and Mayser Industry & Mechanical Engineering [12].
According ISS Safety are shearing and crushing edges at automatically driven devices bear a substantial risk of injury for persons. To protect against these risks, Safety Bumpers are employed. Safety Bumpers, when activated, will immediately switch off the power source. An additional control unit is not required. The signal is fed to the existing E-stop relay unit. The Safety Bumpers comprise a rubber-foam cushion, in which a contact chain is embedded. Normally, the rubber-foam cushion is provided with a high abrasion resistant, elastic polyurethane surface. For applications in harsh conditions, additional materials and solutions can be supplied.
Safety components
2017.TEL.8176 22
3.2.2. Safety edges sensors
Safety Edges, that are produced by Zippswitch Products [20], are industrial grade, pressure sensitive, safety switches. Safety Edges are mounted to moving parts of an industrial machine that may cause a hazardous work environment. The purpose of a Safety Edge is to protect personnel and machinery. The Safety Edge is used as ‘pinch point’ and ‘collision’ safety sensors. They are designed to detect contact of an object and to emit a signal to the machine controls in response. When activated, the Safety Edges acts as an emergency stop that removes power from the machine. Safety Edges are safety devices which are used as an integral part of the machine safety circuit.
figure 17: Example of a safety edge sensor [21]
The safety edges produced by Bircher Reglomat [21] have the following working principle: When a force is applied to the rubber profile (figure 17), the profile makes contact with the conductive elements in the contact strip (figure 18). An electrical signal is generated, this signal will stop the AGV immediately to achieve safety.
The following formulas are used to determine the stopping distance and the overtravel. Stopping distance of parts (AGV):
10. 𝑠1= 1 2∗ 𝑣 ∗ 𝑡
Minimum overtravel of the safety edge: 11. 𝑠 = 𝑠1∗ 𝑐
Safety components 2017.TEL.8176 23 𝑠1= 𝑠𝑡𝑜𝑝𝑝𝑖𝑛𝑔 𝑑𝑖𝑠𝑡𝑎𝑛𝑐𝑒 [𝑚] 𝑠 = 𝑜𝑣𝑒𝑟𝑡𝑟𝑎𝑣𝑒𝑙 [𝑚] 𝑣 = 𝑠𝑝𝑒𝑒𝑑 𝑜𝑓 𝑡ℎ𝑒 𝑝𝑎𝑟𝑡𝑠 (𝐴𝐺𝑉) [𝑚𝑚 𝑠 ] 𝑡 = 𝑜𝑣𝑒𝑟𝑡𝑟𝑎𝑣𝑒𝑙 𝑡𝑖𝑚𝑒 𝑜𝑓 𝑡ℎ𝑒 𝑒𝑛𝑡𝑖𝑟𝑒 𝑠𝑦𝑠𝑡𝑒𝑚 (𝐴𝐺𝑉 + 𝑠𝑎𝑓𝑒𝑡𝑦 𝑒𝑑𝑔𝑒)[𝑠] 𝑐 = 𝑠𝑎𝑓𝑒𝑡𝑦 𝑓𝑎𝑐𝑡𝑜𝑟 (𝑎𝑡 𝑙𝑒𝑎𝑠𝑡 1.2)
figure 18: Working of safety edge [21]
3.3. Emergency stop button
Every AGV is equipped with an emergency stop button. If the button is activated, the AGV will immediately stop. In a system with AGVs, the motion capable equipment will become inactive [22].
Safety control
2017.TEL.8176 24
4. Safety control
4.1. Collision avoidance control
Various collision avoidance techniques have been proposed in literature. Different collision avoidance techniques and or methods are mentioned below. The following methods will be discussed:
• Petri net method
• Game theory method combined with modern control theory • Time window method
• Dijkstra’s algorithms • Hulls
• Ultra-wide band technology
4.1.1. Petri net method
The basic explanation of the Petri net method is retrieved from David A. Tanzer [24].
A Petri net is a graph with two kinds of nodes: species and transitions. The net is populated with a collection of ‘tokens’ that represent individual entities. Each token is attached to one of the species nodes, this attachment indicates the type of the token. A species node is like a container that holds all the tokens of a given type. The transitions represent conversion reactions between the tokens. Each transition is connected to a collection of input species-containers, and to a collection of output containers. When it ‘fires’, it removes one token from each input container, and deposits one token to each output container. Here a simple example for a simplistic model of the formation and dissociation of H2O molecules. The circles are for species, and the boxes are for transitions:
Safety control
2017.TEL.8176 25
The transition takes in two H tokens and one O token, and outputs one H2O token. The reverse transition is split, which takes in one H2O, and outputs two H’s and one O.
This difference in scale leads to a qualitative difference in the modelling. With small population sizes, the stochastic effects will predominate, but with large populations, a continuous, deterministic, average-based approximation can be used.
The mathematical explanation of the Petri Net Method is described by Juntao Li et al. [9] Petri Net method was presented by Carl Adam Petri in 1960. The method is suitable to describe asynchronous and concurrent computer system model, and describing the phenomenon of the
concurrency in a physical light, and it can describe the dynamic performance of the system greatly. At present, the method has been used to study collision avoidance of AGV and system scheduling. Generally, the steps of applying Petri Net method to study the issue are the next;
• construct n-tuple for the problem
• construct the model of Petri net which can describe the problem • apply the relevant theory to analyze and solve the problem.
In the research Maria Pia Fanti [25], the application of Petri net approach to this problem has achieved very gratifying results. In 2002, Maria used colored Petri net method to study AGV collision avoidance of multi-vehicle AGVs, in the case of multi-vehicle AGVs, color and token in Petri net should be one-to-one correspondence between AGVs. If the color is not token, then it will be difficult to distinguish different AGVs while researching the model, and makes it difficult to grasp the whole model, so to grasp the overall model, this paper will color tokens in different colors to distinguish the different corresponding AGV, color the token pass libraries and the transition paths on the corresponding color, thereby forming a colored Petri nets.
Coloring Petri Net (CPN) is defined below: CPN is a directed graph, which can be expressed as a five-tuple, among which:
12. 𝐶𝑃𝑁 = {𝑃, 𝑇, 𝐶, 𝐼, 𝑂}
• P and T are expressed as a collection of pass libraries and transition paths, their definitions are the same with basic PN
• C is expressed as a collection of the colors of pass libraries and transition paths, in particular: 𝑝𝑖′𝑠 collection of the colors of pass libraries:
13. 𝐶(𝑝𝑖) = {𝑎𝑖.1, … , 𝑎𝑖, 𝑢𝑖}, 𝑢𝑖= |𝐶(𝑝𝑖)|, 𝑖 = 1, … , 𝑛
Safety control
2017.TEL.8176 26
14. 𝐶(𝑡𝑖) = {𝑏𝑖.1, … , 𝑏𝑖, 𝑣𝑖}, 𝑣𝑖= |𝐶(𝑡𝑖)|, 𝑖 = 1, … , 𝑚
• 𝐼(𝑝, 𝑡) expresses from the library p to transition t of input mapping (function):
𝐶(𝑝) × 𝐶(𝑡) → 𝑁 (Non-negative integer), which corresponding coloring directed arc from p to t, and the 𝐼(𝑝, 𝑡) here is the matrix;
• 𝑂(𝑝, 𝑡) expresses from the transition t to library p of output mapping (function):
𝐶(𝑡) × 𝐶(𝑝) → 𝑁 (Non-negative integer), which corresponding coloring directed arc from t to p, and the 𝑂(𝑝, 𝑡) here is the matrix;
If you consider the time factor to the colored Petri net method, it will constitute a timed colored Petri net (CTPN). If you place the library as a node, and the change as a side, then the essence of Petri net is a figure and of a general nature, of course, it also has a lot of qualities.
In recent years, Petri net in the AGV collision avoidance program areas focused on the application of colored Petri net approach. Petri net is mainly focused on the case of a deadlock and the AGV collision prevention, but slightly less on real-time collision avoidance.
4.1.2. Game theory combined with modern control theory
The game theory method is described by Juntao Li et al. (Li, Tao, & Xiang, 2016)
Game theory is the study of the nature of phenomena or competitive fight mathematical theories and methods. Game theory is widely used in economic and industry. Game theory can solve a class of problems under the background of competition well in industry, such as the issue of collision avoidance planning of multi-vehicle AGVs. Fundamentally it can be understood as different AGVs compete for the same resources in the same time. The game theory is put model about multi-vehicle AGVs at road intersections. The background to the issue as shown below:
Safety control
2017.TEL.8176 27
figure 21: The schematic diagram of collision avoidance simulation situations [9]
Make a model of AGV A and AGV B by using state-space method, the state-space representation is a mathematical model of a physical system as a set of input, output and state variables related by first-order differential equation:
15. 𝑥 = [𝑥𝑥𝑎𝑎 𝑟 ] = [ 𝑥1 𝑥2 𝑥3 ] ∈ 𝑅3 𝑥𝑎= 𝑡ℎ𝑒 𝑑𝑖𝑠𝑡𝑎𝑛𝑐𝑒 𝑓𝑟𝑜𝑚 𝐴𝐺𝑉 𝐴 𝑡𝑜 𝑁 𝑥𝑏 = 𝑡ℎ𝑒 𝑑𝑖𝑠𝑡𝑎𝑛𝑐𝑒 𝑓𝑟𝑜𝑚 𝐴𝐺𝑉 𝐵 𝑡𝑜 𝑁 16. 𝑟 = 𝑥𝑏− 𝑥𝑎
It can also be expressed as: 17. 𝑥 = [00 10 00 0 −1 0 ] 𝑥 + [ 0 1 0 ] 𝑥𝑎+ [ 0 0 1 ] 𝑥𝑏= 𝐴𝑥 + 𝐵𝑢 + 𝐷𝑑
Game theory is applied to the issue of collision avoidance, and the issue as a game of two persons zero sum, in the process of collision avoidance of two AGVs, the strategy must be one AGV pass through and the other must stop and wait. It can be understood as the victory of the competition when one gets the benefit and the loser lost the benefit.
The cost function can be expressed as: 18. 𝐽(𝑥0, 𝑢, 𝑑) = |𝑥
Safety control
2017.TEL.8176 28
Solving the collision problem would be translated to find out one (𝑢, 𝑑) of the following conditions 19. 𝐽(𝑥0, 𝑢, 𝑑) > 𝑐 = |𝑥
3(𝑡)|
The (𝑢, 𝑑) which meeting the conditions can be expressed as the saddle points of the cost function. Modern control theory has the advantage of being able to accurately model for the describing physical system. And the advantage of game theory is that it can make rational analysis of the competitive background of the described phenomenon. When the combination of them, this problem can be solved. When the collision problem for two-person zero sum game must be solved, one party will benefit and the other will lose interest. This approach would also undermine the interoperability of systems, so the issue of collision avoidance is Nash equilibrium. The Nash equilibrium [26] is a concept of game theory where the optimal outcome of a game is one where no player has an incentive to deviate from his chosen strategy after considering an opponent's choice. Overall, an individual can receive no incremental benefit from changing actions, assuming other players remain constant in their strategies. A game may have multiple Nash Equilibria or none.
4.1.3. Time window method
The time window method is retrieved from Juntao Li et al. [9]
A time window is a collection that constitutes a specific time of the execution of an AGV. The time window is divided into a reserved time window and a free time window. The reserved time window is the time interval that the AGV occupies in one of the nodes, a free time window is the time window that can schedule the other AGV within the reserved time window. Spatial feasibility is the availability of a path physically between two nodes, time feasibility is that the AGV can leave the node within the beginning of the next free time. The time window method is mainly detecting whether a collision will occur between the AGVs by checking the spatial and temporal feasibility.
The next, the analysis of the two conflicts which combined with the time window is shown in the following illustration. AGV1 moves from node i to node j, AGV2 moves from node j to node i, this kind of conflicts will be detected by comparing the order of two nodes, when it is anti-sequence, there is a collision problem.
Safety control
2017.TEL.8176 29
figure 23: The schematic diagram of opposite of conflict time window [9]
As shown in the following illustration (figure 24), in the case of two AGV are in the uniform speed, after AGV1 arriving at the node j and leaving directly, AGV2 arrives at this time, then there will be no conflict between the two AGVs. But if AGV1 turns around in the nod, the reserved time window, which AGV1 occupying in this node, will be lengthened. If AGV2 moves uniformly in accordance with the original planning path, then there will be a collision course between AGV1 and AGV2.
figure 24: The schematic diagram of the catching up of conflict situations [9]
Safety control
2017.TEL.8176 30
4.1.4. Dijkstra’s Algorithm
Retrieved from Mark Duinkerken [27], Dijkstra’s algorithm is developed in 1959 by a Dutch computer scientist Edsger Wybe Dijkstra. His algorithm was first intended to demonstrate the use of computers. Later this algorithm was also used to find shortest paths between different nodes in a graph. This
algorithm is later extended to many variants. A small explanation of the working of the algorithm is stated below Start with the origin as solved node
• For the nth iteration:
o from each solved node, find the nearest unsolved node that can be reached o determine the unsolved node with the shortest distance from origin
o mark the found node as solved; keep record of the added connection • Stop when new solved node is the destination
An example of the shortest path found by the Dijkstra algorithm is stated below, this example is retrieved from Wang Shu-Xi [28]:
figure 26: Shortest path calculation according Dijkstra’s algorithm [28]
In the table on the next page is the derivation of the shortest path determined, there are two solutions of the shortest path in this example.
Safety control
2017.TEL.8176 31
table 1: Iteration table shortest path algorithm [28]
Iteration Solved Nodes Connection Length Solved
1 V1 V1>V2 1 V2 2 V1, V2 V1>V4 V2>V3 3 1 V3 3 V1, V2, V3 V2>V5 V3>V4 V1>V4 4 3 3 V4 4 V1, V2, V3, V4 V2>V5 V3>V6 V3>V7 V4>V5 4 5 5 7 V5 5 V1, V2, V3, V4, V5 V5>V6 V3>V6 V2>V7 5 5 5 V6 V7 6 V1, V2, V3, V4, V5, V6, V7 V6>V8 6 V8 V1, V2, V3, V4, V5, V6, V7, V8
There are two solution paths here, now started from end to start: 𝑉8 > 𝑉6 > 𝑉5 > 𝑉2 > 𝑉1
𝑉8 > 𝑉6 > 𝑉3 > 𝑉2 > 𝑉1
Both paths have a length of six, they are both shortest paths. Concluded, from start to end:
𝑉1 > 𝑉2 > 𝑉5 > 𝑉6 > 𝑉8 𝑉1 > 𝑉2 > 𝑉3 > 𝑉6 > 𝑉8
Safety control
2017.TEL.8176 32
In addition to the general Dijkstra algorithm, according to Atique Shaikh et al. [29] should the AGVs that are available, when they get any command, be allocated and should be send to the destination using the shortest route. If all the AGVs are engaged then depending upon the nearest distance of the AGVs, they should be called using the shortest path and then the AGV should be allocated the load and it should reach the destination using the shortest route. It is well known that the selection of a certain route and a time schedule influence the overall intelligent warehouse system performance. Therefore, one of the main purposes of AGV routing system is to minimize the time waste in cargo transportation.
Earlier in static routing, all the information such as position, and cargo demand were known, route calculation does not change which do not consider collision avoidance procedures which affect the system drastically due to deadlock and traffic jams. So, the algorithm will first check the availability or the nearest availability of the AGVs for the allotment and then try to find the real time most efficient route for all AGV based on the several environment data and warehouse priorities information.
The goal is to find an approach for developing an approach for generating a path planning algorithm and testing the algorithm efficiency in different working conditions to find the shortest path. It will also take care if any obstacle comes in the path, so it will recalculate a new path and change the path.
Based on the 3D model of the warehouse a 2D map will be generated. From the 2D map, all the features will be labeled as node and given a number. Each node will have its coordinates in the rectangular coordinate representing x and y addresses in meters. The AGV control point will calculate the shortest path in the routing task and send it to the nearest AGV.
figure 27: Representation of a 2D proposed warehouse [29]
Safety control
2017.TEL.8176 33
• Check the Load for selecting the numbers of AGV • If Load > 70kg, select number of AGVs per 70kg
• Check the destination and look for any presence of obstacle in the path then calculate the route to the destination from the loading point avoiding the obstacle using Dijkstra’s Algorithm. • Check for the availability of AGVs at loading point and assign the available AGVs
• If no AGV are available at loading point, calculate the shortest distance of all the AGVs to return to the loading point and assign the load to the nearest AGV
• At the next command. Calculate the shortest distance for the remaining AGV and assign the load from them
• Repeat the cycle for the next load.
The Task undertaken by AGV is represented by the flowchart as follows:
figure 28: Flowchart representation of the routing task [29]
This is a short overview of the working of AGVs in an operating system according the shortest path algorithm developed by Dijkstra, in this case a 2D model of a proposed warehouse is used as example (figure 27).
4.1.5. Collision avoidance using hulls
Safety control
2017.TEL.8176 34
There are transport and AGV agents, the choice to let each AGV be controlled by an AGV agent is obvious. Transports have to be handled in negotiation with different AGVs, therefore transport agents are defined. Both types of agents share a common architectural structure, they have different internal structures that provide the agents with different capabilities.
• Transport Agent: Each transport in the system is represented by a transport agent. A transport agent is responsible for assigning the transport to an AGV and reporting the status and completion of the transport to the client that has requested the transport. Transport agents are autonomous entities that interact with AGV agents to find suitable AGVs to execute the
transports. Transport agents reside at the Transport Base, i.e. a dedicated computer located in the warehouse.
• AGV Agent: Each AGV in the system is controlled by an AGV agent. The AGV agent is responsible for obtaining and handling transports, and ensuring that the AGV gets maintenance on time. As such, an AGV becomes an autonomous entity that can take advantage of
opportunities that occur in its vicinity and that can enter and leave the system without interrupting the rest of the system. AGV agents are deployed on their associated AGVs.
AGV agents mark the path they are going to drive in their local virtual environment using hulls. The hull of an AGV is the physical area the AGV occupies. A series of hulls describe the physical area an AGV occupies along a certain path. If the area is not marked by other hulls (the AGV’s own hulls do not intersect with others), the AGV can move along and drive over the reserved path. In case of a conflict, the involved local virtual environments use the priorities of the transported loads and the vehicles to determine which AGV can move on. AGV agents monitor the local virtual environment and only instruct the AGV to move on when they are allowed. Afterwards, the AGV agents re-move the markings in the local virtual
environment.
The next section is an addition to the section above and retrieved from Elena Cardarelli et. al [31]. Since AGVs are constrained to move on fixed paths, they are not allowed to deviate from the roadmap; therefore, in the presence of unforeseen obstacles it is necessary to find an alternative path on the roadmap, which is not always possible as, for instance, in the frequent case of mono-directional roads. In this situation, the AGV is forced to stop, until the obstacle has been removed. This lack of flexibility strongly impacts on productivity and huge traffic problems can be generated, especially in the presence of multiple AGVs, see figure 29.
Safety control
2017.TEL.8176 35
figure 29: Multiple stuck automated guided vehicles [31]
Because the ability of the presence of human workers and manually driven forklifts in warehouses where AGVs are operating, sensors are installed onboard of each AGV, providing a reliable and robust obstacle detection in the vehicle surrounding area. However, these sensors are not capable of classifying the detected objects, preventing the possibility of differentiating the motion control of the AGV according to the type of obstacle in its sensing range. While static obstacles could be easily passed, without having any negative impact on the overall safety of the system, conversely, if a human is crossing the sensing range of an AGV, the only safe procedure is to avoid any movement. However, these high-level decision strategies could not be implemented, because some sensors do not allow to distinguish between humans and other kind of obstacles. Moreover, since obstacle detection is limited to the neighborhood of each AGV, global information on the surrounding environment is not available, hence, to avoid collision with unexpected moving objects in critical zones as, for instance, intersections and blind spots, the AGV is forced to reduce its speed.
Safety control
2017.TEL.8176 36
figure 30: Advanced perception system [31]
As a motivating example, consider the scenario in figure 30: in the presence of multiple objects, as well as pedestrians, an AGV is moving on the path of roadmap represented by the orange dashed line. The on-board perception system allows to identify only a limited portion of the obstacles about the AGV (green area in the figure top left). Therefore, based only on local sensing, it is possible to compute an obstacle-free deviation from the roadmap (blue line in the figure top right). However, as represented in the figure bottom left, exploiting the local sensing capabilities could not be sufficient to guarantee collision
avoidance: following the obtained maneuver, the AGV conflicts with a previously occluded pedestrian, not directly visible by the on-board perception system during the computation of the local deviation (top left figure).
The mobility of the AGVs impose highly dynamic operation conditions and inherent distribution of
resources. A typical approach in mainstream software engineering is to support distribution with a suitable middleware. Mobile applications such as an AGV transportation system are characterized by:
• Their need to consider their physical environment
• Their need to deal with dynamics and unexpected events originating from their context The development of the application components of the AGV transportation system are simplified. The middleware encapsulates the tedious management tasks associated with distribution in mobile systems.
