ITMO: A universal method of designing adaptive and energy-efficient robots

Main article: Robots (robotics)

2023: Developing a method to design adaptive and energy efficient robots

ITMO scientists have developed a universal method for designing adaptive and energy-efficient robots that are able to move in unfamiliar space, safely interacting with the environment, objects and people. The proposed method greatly simplifies the creation of robotic devices. The university announced this on May 31, 2023.

The method proposed by scientists is a universal framework based on algorithms. It allows you to design mobile robots (walking, jumping), most of the required dynamics of which are ensured by their physical execution: mass distribution, elasticity, mechanical bonds. Such systems with "smart" mechanics can work longer without recharging, are able to adapt to new conditions and move over uneven surfaces. Scientists have already applied the method to develop a gripping device, jumping and walking robots, as well as an exoskeleton.

As soon as walking robots appeared, working not from an outlet, but batteries, the developers faced the task of increasing their energy efficiency. Devices created according to the principles of industrial robotics can outwardly resemble humans or animals. They are created from heavy parts and drives, they also depend heavily on vision and sensors - because of this, mobile robots consume too much energy and cannot move through rough terrain. Our method is based on another approach: it allows you to design, maybe not such accurate and fast systems with high load capacity, but adaptive and energy efficient, able to interact with the environment. For example, these are adaptive brushes, blindly exciting objects, or walking robots that move in space, even if a three-dimensional map of the environment is built only partially or with an error, said Ivan Borisov, associate professor of control systems and robotics at ITMO.

The method consists of three stages. In the first, the framework solves the problem of open kinematics: how many consecutive links should consist of a mechanism, what length they should be, what trajectory with their help should be built in order to achieve the desired behavior and performance. On the second, additional links are added to the structure, they turn it into a more complex closed kinematics. Closed kinematics allows you to get a mechanism with less inertia and get rid of unnecessary drives. And finally, at the third stage, the framework introduces elastic elements into the resulting system and calculates where to place the drive that will drive the robot. This step reduces the amount of energy consumed.

By entering the minimum number of parameters into the framework, the engineer receives a device project for his tasks. All parameters are calculated in it, geometry is optimized, weight and elasticity are correctly distributed. All this helps to achieve a design where engines and microcontrollers do not provide the entire movement process, but only correct behavior. Moreover, we managed to reduce the requirements for the number, power and, accordingly, the size of the drives - the most expensive and heavy element. We tested the method on different mechanisms, our last prototype of a jumping robot showed a decrease in drive power by 4 times compared to classical mechanisms, - said Kirill Nasonov, engineer of the Faculty of Control Systems and Robotics ITMO.

According to the authors of the project, in the near future the method will be improved to a software product that fully automates the process of designing lever mechanisms for mechatronic and robotic systems.

At the input, the engineer will need to set the minimum number of characteristics, and at the output of the algorithm, choose the most suitable from a dozen options ranked by efficiency. If suddenly none of them comes up, the specialist will be able to specify his request in order to get a new selection: set more parameters or make changes to open source, - said Dmitry Ivolga, engineer at the Faculty of Control and Robotics Systems at ITMO.