As it explained in my last blog post, I am creating a simple AI agent to improve the driving conditions of a semi-intelligent car, using the concepts of classifier systems. Each true or wrong statement can be shown with a chromosome of numbers. If-part represents the current condition of the environment, and Then-part recommends the action to act upon the environment, based on the If-part of the chromosome. Note that I aim to find and keep good rules (true rules) and withdraw bad rules (wrong rules) in the training process. I am defining a new term, strength (St), for each chromosome that indicates how accurate the chromosome can act upon the environment based on its If-part.
Later, during the learning process, the strength of a rule will be modified based on the accuracy of its Then-part, i.e. the strength of a rule will be increased if it has a right Then-part and vice versa. For this example, the intelligent agent is a set of 11 rules (The number of rules in real examples range from couples of hundreds to thousands), shown in the table below.
The following two examples (It can be replaced by historical data) are used to train the system (note that examples must have a value for ALL parameters in the If and Then-parts):
1) The windshield is wet and the distance from the car ahead is too close and the velocity is low and the engine is on and the lighting is sufficient and the temperature is high ð only wipers are turned on (no other action is taken)
The example is coded as: 100111:1####
2) The windshield is dry and the distance from the car ahead is fine and the velocity is high and the engine is on and the lighting is sufficient and the temperature high ð only doors are locked (no other action taken)
The example is coded as: 011111: ####1
In my next blog post, I will explain how to train the Intelligence System based on two examples mentioned above. For the full list of my post, please check my LinkedIn.
Concluding my last two posts, Artificial Intelligence Systems in Easy Word, there are four units and two phases for the majority of artificial intelligence systems, shown in the table below.
There are different ways of creating intelligent agents such as Neural Networks, Fuzzy Logic, and Classifier Systems. Sometime researchers combine these methods to create better agents for their specific question or they are using genetic algorithm to improve the intelligence of their system. In these series of blog posts, I don’t discuss these methods of creating intelligent agents, but help you to understand the requirements of creating artificial intelligence system and finding questions for this solution. Yes, finding some question to solve them with AI solution.
Here in these series of my blog post, I am creating a simple AI agent to improve the driving conditions of a semi-intelligent car, using the concepts of classifier systems. This example is simplified version of eXtended Classifier System (XCS) example from my notes from my very first AI class with Dr. Masoud Shariat Panahi, Associate professor of University of Tehran. In this example, we develop an intelligent agent to improve the driving conditions of a passenger car.
This car is equipped with six sensors in order to watch the environment and five actuators to make an action. Here’s the list of sensors and actuators.
Each true or wrong statement can be shown with a chromosome of numbers. In other words, we can standardize our statements by using chromosome filled with 0/1/# (# for a sensor stands for the condition of 0 or 1; and for an actuator stands for do nothing). For example, the statement “If windshield is wet, then turns on the wipers” is represented by:
Or even we can simplify it to 1#####:1####.
Out intelligent agent is a set of thousands rules like 1#####:1#### that can be true, false, or partially true. These rules are generated randomly. In my next blog post, I will train the system and explain how it is possible to create intelligent agent.
The core of any intelligent agent is continuous learning. Our brain tried thousands of ways that does not keep us stable when walk, but only one (or couples) of those thousands ways helped us to learn how to walk. Note that it’s continuous learning, not just learning. As we are getting older, our brain learn how to keep the stability to handle taller and heavier person. When we are learning how to ski, our brains are trying thousands ways of falling down. Each try to keep our stability is a rule in our brain. Each rule could be as simple as an if-then statement in our mind like “when your eyes detect you are getting close to uphill, you should bend at the hip” or it can be more complicated rule when we start running or skiing. Each rule can be a true or false statement in our mind. Our brains are trying thousands of rules; then brain strengthens good rules (ways that keeps our stability) and weakens bad rules (rules that leads to fall).
The beauty of any intelligent system comes from trying thousands of random rules and finding good rules. In practice, when intelligent agent acts on an environment, it gets feedback from the environment to update the strength of applied rule. In walking example, falling down is a negative feedback in our brain, and keeping stability is a positive one. But what if we want to train an artificial intelligent agent to keep the robot stability? Or to predict the stock market?
Most of artificial intelligent agents need to learn from another intelligent agent, like when your parents teach you how to speak and then you become a poet. Artificial agents are programmed by programmer and they are trained (finding good rules or true stamen) by training data. This introduction leads us to couples of important concepts in AI methods. (Photo Credit)
- Training data: training data is a set of data used to discover potentially predictive relationships. In stock market, training data could be historical data of stock market. When you are training a robot for stability, you training data are any data that keeps your stability in any certain condition.
- Machine learning: machine learning is a program that provides computers with the ability to learn from training data. Machine learning focuses on the development of computer programs that can change when exposed to new data.
- Supervised learning: Supervised learning is the machine learning task of inferring a function from training data.
Do you want to learn more? In my next blog post, I will cover more concepts of artificial intelligent systems and create a simple intelligent agent to improve the driving conditions of a passenger car.
From the time I registered as a MBA student at University of Nevada, I have been talking to my colleges about Artificial Intelligence (AI) and its application in economy, finance, pricing, and literally for everything. We were sometimes struggling on the privacy issues of intelligence systems and weather these intelligence systems can overcome the human’s intelligence. I, as the only engineering student in my MBA program who has worked with intelligent systems, was always thinking of defining artificial intelligence in easy words; something beyond just definition, something practical. In this and couple of more upcoming blog posts, I want to bring up a very simple example of artificial intelligence application to cover all concepts.
An intelligent agent is a core of intelligent system. Intelligent agent defined as an autonomous entity that observes the environment and acts upon an environment (i.e. it is an agent) and directs its activity towards achieving a goal or goals. The most familiar and the most complicated intelligent agent is our brain. One of our first brain’s goal was learning to walk; our brain was and still is observing environment through vision sensors, which are our eyes, and acts upon an environment by using our muscles as an actuators.
The same thing happens with artificial intelligent agents. There is always a device or medium that an intelligent agent observes the environment through that, and also, a device or medium that the intelligent agent impacts the environment. Obviously, this ongoing process happens to gain advantage or to achieve the goal. If I compare human’s brain as an intelligent agent with other artificial intelligent agents, I would get the table below.
||observes the environment through
||acts upon an environment through
||Bipedal actuators (motors)
||To keep the robot’s stability
||Historical stock market data
||Selling or buying stocks
||Gain profit by selling and buying stocks
|Amazon’s supply chain
||Historical buyer’s shopping behavior
||Predicting buyer’s shopping in future and sore goods in warehouse near end-user customers
||Lowering shipment cost
Questions that would come up after this comparison are:
- What makes our brain intelligent?
- Can we create an artificial agent that works like brain?
In my next blog post I will answer these two questions as well as covering more concepts and examples.
Hydraulic machines are machinery and tools that use liquid fluid power to do works. Hydraulic fluid is transmitted throughout the machine to various hydraulic motors and hydraulic cylinders and is controlled by control valves and distributed through hoses and tubes. The beauty of hydraulic systems is due to the capability of creating large amount of power by using small tubes and flexible hoses. Hydraulic machines are normally filled with oil as liquid fluid to transmit power. In water hydraulic systems, water is used instead of oil as pressure medium. In this post, I highlighted the advantage of using water instead of oil as pressure medium.
Although water hydraulic systems are more expensive than oil ones, there are several advantages that ingratiate using of water instead of oil in hydraulic systems. The first and the most advantage of water hydraulics is their fire resistance. Using water for hydraulic systems is specially a great safety advantage for those industries that they need to have high voltage electricity next to their hydraulic systems. Moreover, water is more available, cheaper, and requires much less purification than oil. Considering environmental concerns, water hydraulics systems provide an environmentally friendly and less-costly to recycle alternative compare to oil-based hydraulics systems.
Water was the first medium ever used in simple hydraulics systems over 2000 years ago. However, the use of water as a pressure medium was declined in 20th century due to following reasons:
- Corrosion in piping and components due to the micro-organism in water
- Freezing of water under 0 °C
- Water based hydraulic systems need more sealing for the internal and external leakage of water due to lower viscosity
- The severity to control water hammer effect even though it makes the water hydraulic system more accurate
- Poor lubrication of water limits its application for high load and high speed applications.
Many researches developed different solutions to overcome shortcoming of water hydraulic systems. Is water hydraulic a right solution for your business? Please message me here or on my LinkedIn to find potential use of water-based hydraulic systems for your business.
Hydraulics & Pneumatics
Water Hydraulics – Theory and Applications by Gary W. Krutz and Patrick S. K. Chua
Radio Frequency Identification (RFID) is the use of memory chip along with antenna to save and transmit information. RFID is a tag that is attached to a product, animal, or even a person for the purpose of identification via radio waves. Unlike barcode, traditional identification
technology that barcode reader can read the barcode one at a time, RFID reader reads hundreds of tags within fraction of second. From the first use of RFID for tracking the chassis carriers of General Motors, many companies considered this new technology to develop their organization. Photo Credit
Nowadays, RFID technology is used for wide range of application, from pets’ identification, to anti-theft systems and supply chain management. RFID chips can be embedded under a dog or a cat’s skin for the permanent identification; or it can be attached to an expensive instrument to scan and track the object after theft. Automatic identification (Auto-ID) is a new technology that uses RFID chips to collect data and store it in a cloud based computer system without direct human involvement. Walmart as the world’s largest retailer have employed this technology to cut costs in order to save money for its customer to live better. In order to track Walmart pallets, RFID tags are attached to pallets. Anytime pallets pass a gate, RFID readers capture the time of departure, destination and continents of each pallet. This data will be stored in clouding system of Walmart, and then supply chain managers of Walmart are capable of accessing all shipment progress by logging into their accounts.
The most serious shortcoming of RFID is privacy. Anyone with an equipped RFID receiver is capable of accessing RFID content and read its data, as an example, this issue gets out of control when military vehicles are scanned by the other side. For the use of supply chain, the implementation of RFID is still expensive. That’s why the decision of RFID implementation is more a business decision rather than technological one. The last issue that limits the wide use of RFIDs is the fact that RFID reader should be enough close to RFID to read data. This especially limits the use of RFID compared to GPS systems for anti-theft purposes.
As it discussed in previous post, soft grippers will be a new generation of grippers because of their universal application for flexible manufacturing. Soft gripper range widely from those that are working with air pressure to electrically actuated ones. We can also classify soft grippers into two groups of elephant-trunks-shaped and octopus-arms-shaped soft grippers. In this post, we will briefly introduce different kinds of soft grippers and their further application.
Beyond the use of air pressure for the simplest form of pneumatic soft grippers that we explained about in our previous post, Eric Brown Lab used the air pressure to fabricate a pneumatic universal robotic gripper based on jamming of granular materials. A griper’s actuator is an elastic bag that is filled with granular matter. When the gripper touches the surface of an object, an evacuation pump, which is connected to the bag, starts evacuating to lock granular materials around the object. The proposed system is capable of picking up objects with wide range of different sizes.
Bristol Robotics Laboratory introduced an elephant trunk-shape gripper that picks up and holds roughly cylindrical objects. This gripper consists of flexible outer shell that is filled with incompressible liquid, and is activated with two quasi-longitudinal cables. These cables are located in one side of the gripper to provide bending movements.
In out lab at Shan Research Group, we designed and implemented a soft gripper, based on air pressure, consist of three soft fingers. Each finger is separately controlled by three electrically activated tendons that capable our gripper with providing more complex motions than any other hand-shaped gripper. To date, our gripper is the first generation of grippers that enables twisting motion, just like human hands. It is completely safe for human and environment to work with the gripper since they are fabricated from body friendly flexible materials for knuckles and joints. More details about this gripper will be available soon. Would you need any further information about soft gripper, please contact me.