Autonomous systems are a reality due to the increasing technological possibilities such as artificial intelligence, new network technologies and smart energy supplies. In practice, we have learned that constructing safe and reliable autonomous systems requires knowledge of the necessary (enabling) technology and of the domain.
Article from Objective 26, 2016
Where in the past people would operate a device on location, it is currently often controlled remotely from a central spot. At container terminals, the crane operator is no longer sitting in the crane, but controls a crane robot from a control room. Traffic centers also fulfill more responsibilities. This trend coincides with the rise of autonomous systems. Due to the ever-increasing automation, machines can often function without an operator. Lots of product lines in factories are completely automated. People are not required to perform boring repetitive or dangerous jobs any more. The machines are functioning 24 hours per day on autopilot, which results in a huge cost reduction and is, in addition, much safer.
The route to autonomy
The term ‘autonomous’ can lead to confusion. The traditional thermostat of the boiler is autonomous, isn't it? No, it is automatic. However, the user determines the conditions: at 07:00 AM I want it to be 18oC. The build-in thermometer measures the temperature of the environment and makes the boiler remain active until the 18oC has been reached. This old-fashion thermostat is rather low on the autonomy ranking; we call this an isolated system. There is a local measure and control system with sensors and actuators (on/off-switch). A modern thermostat ranks a little higher. It is connected to the Internet, can remember your behavior, knows how long it takes for your house to warm up and cool down, measures the outside temperature and works together with other systems such as sunblinds. With this information, it creates the right climate. After a few days, it can adjust to accommodate the right temperature for each moment of the day on its own. This is an autonomous system.
Monitoring with decision support systems
The black ice detection system (gladheidmeldsysteem), that alerts the salt spraying trucks when there is a risk of black ice, is in the monitoring stage. It combines the measured data with the information provided by meteorological institutes and weather forecast models in order to determine whether salt spraying is required or not. However, in case of an alert, the salt spraying coordinator assesses all the information and then decides whether it is necessary to start salt spraying. The coordinator also takes information into account that has not been included in the black ice detection system.
Making decisions requires knowledge of the domain
Systems that work up to and including the monitoring stage can be realized without the producer having any knowledge of the user's situation. The producer can work with an assignment as simple as 'make a measuring system that sends out an alarm when it rises above a certain limit'. The interpretation of the alarm is still in the domain of the client. The client uses substantive knowledge to assess what to do in case of an alarm. However, when you allow a black ice detection system to make the decision of 'starting the salt spraying process', this knowledge will have to be included in the system. The producer then has to know about the domain and therefore knows what is important when it comes to making a decision. What does the system need to do and when? This results in an autonomous system that can think, control and act on its own.
For construction of autonomous systems, you need a specific set of technologies and tools. An autonomous system has to be able to collect, process, transmit and analyze measuring data in order to make a decision based on these aspects. In technology, this means: sensors, communication, calculating power, decision algorithms and artificial intelligence. We refer to it as 'enabling technologies'.
This combination of technique and application is our core competence: we can support the client as an equal partner, in his domain and language, and use that to create a complete solution with system integration. Our added value is the fact that we can combine and assemble systems by using enabling technologies.
A good example of market meeting technology to create a (semi) autonomous system with the support of our enabling technologies is Sense2Grow. This open platform links wireless sensors simply to each other to generate data for controlling and monitoring the business processes. Any party can connect any type of sensor: simple and universal. By bringing it all together in one platform, combining and processing data has become much simpler. From our agricultural domain expertise, we often use Sense2Grow forinLoRa communication, solar cells and cloud processing in order to monitor greenhouses in an optimal manner.
From self-thinking to self-acting
A traffic center can work autonomously and start scenarios by itself and alternate based on the traffic demand. But what to do when an autonomous traffic center allows traffic to congest? Who is responsible: the user, the system or the producer of the system? Hence, we currently still allow an operator to make the decision whether or not scenario A should be activated. The final decision and thus responsibility is still with the user of the traffic center. Presently, this is the reason that autonomy is too much to ask for in many situations and thus clients often opt for a semi-autonomous system, one that can think for itself, but will not act on its own. It provides an advice for the user, who then will make the final decision.
For the acceptance of autonomous systems, demonstrable safety is essential. The user needs the guarantee that the system will keep operating in all situations. He/she does not want to deal with gridlock due to a flaw in the software, or tunnels that have to be closed due to defective safety systems. This demand requires special attention for the working method in order to ensure good quality software and technique (see frame), but also knowledge of specific safety rules for that application.
The tricks of the trade
In order to be able to design a system that makes decisions or offers advice, the producer will have to be familiar with the client's language and understand his/her practical problems. A good system producer nowadays has knowledge of the client's domain so they can communicate on the same level, make informed choices and thus will result in a system that can add value.
In addition, the client increasingly approaches the problem in a functional manner, without having any idea of a technological solution. The client no longer asks for a certain piece of technique. The client requires a function, a practical solution to the problem. How to solve such a problem is up to the producer. The tricks of the trade are in the flexibility of the technical world to be able to adapt to everyday practice and vice versa.