Robot Cooks A Tasty Omelette
Robot Cooks A Tasty Omelette
With enhancements in the field of robotic manipulation, sensing and machine learning, robot cooks are awaited to become prevalent in our kitchens and restaurants. Robotic chefs are envisioned to replicate human skills to lessen the burden of the cooking process. Nevertheless, the potential of robots as a means to improve the dining experience is unrecognised.
By using a machine learning technique, scientists have successfully trained a robot to be effective for extremely biased matters of taste. The study was carried out by a research team from the University of Cambridge in partnership with Beko – a household appliance firm.
Advancement in machine learning – robot cooks:
For several years, futurists, researchers, and science fiction authors have been dreaming for a robot that can cook. With the progression of artificial intelligence systems, commercial businesses have created prototype robot chefs, even though none of these models is shortly available in the market. In addition to this, these models lag much behind their human counterparts to skill.
It is a challenging job to train a robot to prepare and cook food because it has to deal with complex issues in terms of computer vision, sensing, robot manipulation, and human-robot communication, and give a reliable, final output.
Taste too varies from one person to another, cooking appears to be a qualitative task, whereas robots normally do well in quantitative jobs. Taste is not universal and, hence, there are no universal solutions. Contrary to other optimization concerns, unique tools must be designed for robots to make food.
Other research groups taught robots to cook cookies, pancakes, and even pizza, such robot chefs have not been adjusted to handle the different subjective variables involved in the cooking process.
The machine learning technique developed by the team:
For a long time, egg dishes, omelettes particularly, have been considered as a test of culinary skill. Omelet is one of those recipes that are easy to make, but tricky to cook well.
Researchers believed it would be the perfect test to improve the robot chef’s skills and optimize taste, texture, smell, and appearance.
Now, in partnership with Beko, Iida and his associates have adequately trained their robot chef to make an omelette, right from cracking the eggs to plating the completed dish. The research was carried out in the Department of Engineering of the University of Cambridge, using a test kitchen provided by Symphony Group and Beko plc.
The machine learning technique devised by Iida’s research team uses a statistical tool called Bayesian inference to squeeze out the maximum amount of information from a limited number of data samples, which was essential to avoid omelettes overflowing with people-tasters.
But how does one satisfy the robot as a chef? “The omelettes overall tasted great much better than expected!
Experimental Setup:
The experiment is carried in a customised kitchen. An egg cracker, bowl, salt and pepper, pan, electric whisker, and a whisk (used for stirring while heating), are tools that are handled by the robot.
To constrain the tools to a known place, the rig includes clips for the egg cracker, bowl, oil dispenser, electric whisker, and the whisk. Likewise, the initial position for the salt and pepper containers are marked.
A UR5 robot arm from Universal Robotics is utilised as the manipulator for the omelette cooking process. The UR5 is controlled by a python script using a provided API to interact with the UR5 control box.
Through the API(Application Program Interface), instructions to the UR5 such as individual joint angle demands, Cartesian coordinate position demands, and force control can be transmitted.
A simple end-effector has been created and manufactured to manage all the tools in the kitchen setup. It contains two fingers that can move parallel to each other independently through two DC motors.
The fingers are provided with silicone padding better grasp tools with different shapes. A linear potentiometer is attached to each finger which is used for position feedback and to avoid colliding.
The current input to the motor controller is examined from a shunt resistor, which finally indicates the grasping force. This feedback is used to tune the strength of the grip when an object is sensed.
An Arduino Uno was used to see the sensors on the gripper and control the motor through a dual H-bridge circuit. Serial communication enables the Arduino to communicate with the python script controlling the UR5, to regulate the grasping and arm movements.
Robot Cooking:
Omelette Cooking Procedure:
In the omelette cooking procedure, the robot moves through a sequence of events. The robot is reliant on all the tools set up in the known location to start the process but can accommodate slight changes to the tool placement through the gripper robustness and methods such as force feedback when grasping/using the tools. Manually, the eggs are placed in the cracker.
On developing the experimental setup and the sequence for omelette cooking, a list of essential control parameters like salt, pepper, whisking, cooking time, etc. was identified. The heat-level of the hob, the time from placing the pan to heating, and the time from pouring the contents from the bowl until mixing is kept constant.
The problem that researchers have found is the subjectivity of human taste. People do not give complete measures very well and normally give relative measures when it comes to taste.
Therefore, it was crucial to set up a machine learning algorithm, the so-called batch algorithm, so that people-tasters could give information based on comparative assessments, rather than sequential ones.
The results confirm that machine learning can be used to produce quantitative improvements in food optimization. Moreover, this method can be easily extended to several robotic cooks. Further research should be conducted to analyse other optimization methods and their feasibility.
More,
Artificial Intelligence and Robots ignoring Real Human Intelligence and Manpower
image source
- g6: Readsector.com
- g4: Google.com
- g3: Google.com
- Capture: ieeexplore.ieee.org
- g5: Google.com