Robotics have made tremendous progress in recent years, but cars still have many obstacles before densely entering our lives. The journal Science Robotics outlined ten ambitious tasks that need to be resolved to make this a reality. The editors of the magazine conducted an online survey of unresolved problems in robotics and interviewed a group of industry experts.
New materials and assembly diagrams
Robotics begin to move away from the usual engines, gears and sensors, experimenting with elements such as artificial muscles, soft robotics and new assembly methods that combine many functions in one material. But most of the list of these achievements have not yet passed the stage of the demonstration, and it is too early to talk about unification.
Multifunctional materials combine sensitivity, motion, energy collection or storage and allow the design of more efficient robots. But the combination of these properties in one machine will require new approaches that combine micro- and macro-scale assembly techniques. Another promising direction has become materials that can change over time, adapting or recovering, but much more research is needed in this area.
Bi-inspired and biogybrid robots
Nature has already solved a lot of problems over which robots are breaking their heads, so many of them turned to biology in search of inspiration or even include live systems in their robots. However, the reproduction of the mechanical productivity of muscles and the ability of biological systems to self-feed is confronted with "narrow" places in the development.
The field of artificial muscles has already seen significant progress, but their strength, efficiency, density of energy and power require improvement. The introduction of live cells into robots can overcome the difficulties associated with the use of small robots, and also use biological functions such as self-healing and built-in perception, but the introduction of such components is a difficult task. And although the growing "robocopo" helps us learn the secrets of nature, more work needs to be done on how animals have made the transition from pure flight and navigation to multimodal platforms.
Power and energy
Energy storage is a serious stumbling block for mobile robotics. The growing demand for drones, electric vehicles and renewable energy is pushing the progress in the field of batteries, but the fundamental problems remain largely unchanged for many years.
It follows that, in parallel with the development of batteries, there is a need to minimize energy consumption by robots and equip them with new energy sources. Give the robots the opportunity to use the energy of their environment and transmit energy to them wirelessly - these two promising approaches are being actively studied.
Robot Swarm
A swarm of simple robots that assemble in different configurations for a variety of tasks can be a cheap and flexible alternative to large, specialized robots. Small, inexpensive and powerful elements of equipment that allow simple robots to feel their surroundings and communicate, in combination with AI, which can simulate this kind of behavior, already exist in natural swarms.
It is necessary to conduct more work on effective forms of management on different scales - small swarms can be controlled centrally, but larger ones should be more centralized. They also need to be durable and adaptable to the changing conditions of the real world and resistant to intentional or accidental damage. It is also necessary to work more on the swarms of heterogeneous robots with additional capabilities.
Navigation and Intelligence
A key option for using robots is to study places where people can not enter, for example, into deep sea, space or a disaster zone. This means that they need to be skilled at reconnaissance and navigation without maps, often in a chaotic and hostile environment.
The main problems include the creation of systems that can adapt, learn and recover from navigation failures, and are also able to create and recognize new discoveries. This will require a high level of autonomy that will allow robots to track and reconfigure themselves, create a world view from several data sources of varying reliability and accuracy.
AI for robots
Deep training has enabled machines to recognize patterns and patterns at a new level, but this should be linked to simulated reasoning to create adaptable robots that can learn "on the fly."
The key to this will be the creation of an AI that is aware of its own limitations and can learn to learn new things. It is also important to create systems that can quickly learn from limited data, rather than the millions of examples used in in-depth training. Further successes in our understanding of human intelligence will also be needed to solve these problems.
<>
<>
Neurocomputer Interfaces
Neurocomputer interfaces will allow you to stealthily manage advanced robotic prostheses, and also provide a faster and more natural way to transfer instructions to robots or simply help them understand the mental state of a person.
Most modern approaches to measuring brain activity are expensive and clumsy, so we need the development of compact, ergonomic and wireless devices. They should include extensive training, calibration and adaptation, because we can not accurately read brain activity. In addition, it remains to be seen whether they can work better than simple techniques like tracking eye movements or reading muscle signals.
Social interaction
If robots want to enter the human environment, they will need to learn how to communicate with people. This is difficult because we do not have many well-defined models of people's behavior and we tend to underestimate the complexity of what seems to us natural.
Social robots will have to be able to perceive the smallest social signals, such as facial expressions or intonation, understand the cultural and social context in which they work, and model the mental states of people they interact with, adapting their relationships in the short term and designing long-term relationships.
Medical robots
Medicine is one of the areas in which robots can make a significant impact in the near future. Devices that complement the surgeon's capabilities are already used on a daily basis, but we still can not give them full autonomy because of high rates and risks.
Autonomous assistants in the person of robots will need to learn to recognize human anatomy in different contexts and use situational awareness and voice commands to understand what is required of them. In surgery, autonomous robots can perform routine operations, freeing the surgeon for more subtle and important work.
Microbots working in the human body, too, promise a lot, but are in an embryonic stage of their development.
Ethics and safety of robots
As we overcome current tasks and integrate robots into our lives, we face new ethical issues. Most importantly, we can become excessively dependent on robots.
This can lead to the fact that people will get rid of certain skills and abilities and will not be able to take the reins in the event of failure of the robotic system. We can ultimately delegate tasks that, for ethical reasons, are unpleasant for people, and dump everything onto autonomous systems.
