Last updated February 8, 2018 at 10:38 am
The robot revolution is coming, here’s what needs to be tackled first.
Robotics will be one of the biggest changes to the way that we live and operate in the future, and it’s rapidly approaching. With the power to completely revolutionise science, medicine, manufacturing, exploration and our day to day lives, the next decade will be an incredibly exciting time to live through.
However, there are a number of challenges we will need to overcome first, technically, socially and ethically.
The journal Science Robotics ran a public survey of the unsolved challenges in robotics. An invited panel of experts then sifted through the responses to come up with 10 grand challenges that the field will encounter in the foreseeable future.
Seven of these challenges are technologies that will have an impact on all applications of robotics – including creating new materials and manufacturing methods, developing cost-effective and long-lasting power sources for mobile robots, and programming artificial intelligence to perform deep moral and social reasoning about real-world problems.
Two challenges represent key example applications of robotics – medical and social – which involve extensive human interaction and encounter their own unique sets of issues in sensing, perception and intelligence.
The final challenge is addressing ethics and security issues of integrating robots into society.
1. New materials and fabrication
The next generation of robots will need to be multifunctional, power-efficient, compliant and autonomous in ways similar to a biological organism.
While gears and electromechanical actuators are fundamental to the operation of many robots today, in the future new materials and ways of controlling movement will be required. One approach taken by labs around the world are artificial muscles, while others are thinking about advanced manufacturing and assembly strategies.
Artificial muscles are more or less exactly what they sound like, materials which act in a way which mimics biological muscle. However, using softer materials to get this pliability comes with the trade-off of lower strength compared to rigid construction. They operate through the shrinkage or expansion of the material, however that does in turn limit their strength. In an effort to solve this problem, labs are already working on ways to make them stronger.
One other advantage of soft construction and artificial muscles is the potential to make them self-healing in the case of damage.
One other advance which is being worked on at the moment is looking at the very construction of robotics themselves. Rather than a nuts-and-bolts approach to assembly, the next generation of robotics will need to incorporate dissimilar materials in ways they can work together. For example, rigid and soft materials, or conductive and dielectric materials may need to be layered together, or incorporated into a single body, the same way our bodies layer and integrate tissues with different properties (eg muscles and bone).
An advantage of this approach is being able to layer sensors and actuators in a way where they can work together more effectively and efficiently.
2. Bioinspired and biohybrid robots
More and more, robotics engineers are looking to nature for inspiration when it comes to design rules to build a robot that performs like a natural system.
One step even further than taking that inspiration is a biohybrid, which incorporates biological material in its design.
The biggest challenges that need to be overcome include developing a battery that can match a metabolic energy source of a living organism, muscle-like actuators, self-healing materials, and autonomy in different environments.
The biggest bottlenecks to see true bioinspired/biohybrid robots exist in the movement and energy sources.
Electromagnetic motors are useful for large robots, but inefficient when scaled down or in soft systems. Similarly, artificial muscle has the limitations above.
No battery can yet match a biological metabolism for energy generation, which likewise limits their size. However if artificial muscles can be developed further, scaling robotic sizes down may minimise this limitation.
3. Power and Energy
Like all electric systems, power sources are one of the major limitations for robotics.
In practice, the usefulness of a mobile, autonomous robot is largely dictated by its battery’s power, size and weight. A battery with long longevity will be large and heavy, but a small battery will have a limited lifespan.
To minimise these limitations work is ongoing into making the components of a robot more power efficient.
However, another approach could be to make an autonomous robot able to extract energy from its surroundings, such as light, vibrations, and mechanical movement.
Battery technology will also likely need to improve beyond the nickel-metal hydride and lithium ion options currently available. Research is already ongoing into next generation options such as fuel cells and supercapacitors.
Making these energy harvesting and storage options stable in different environments will also be a challenge. Deep sea exploration robots require compact, stable and high-energy density batteries. A flying robot would require a lightweight, tolerant to temperature solutions. As a result, it is unlikely one solution will cover all applications, and each will need to be developed individually.
4. Robot swarms
Robot swarms are a different take on tackling a task requiring a large robot – allowing simpler, less expensive modular robotic units to be configured into a team.
It is an approach very much inspired by nature, where a single animal can’t achieve a goal by itself, so instead coordinate with others to complete the task.
This approach however not only takes coordination with others, but also sensing the environment, the other animals around it, and communication within the team, while in effect acting independently on its role in the task.
As sensors, processors, and communication devices get cheaper and increase in performance, with a reduction in size, the development of swarm robot strategies will increase over the next decade.
5. Navigation and exploration
Robotics has moved ahead in leaps and bounds in navigation and exploration, with autonomous robots advancing rapidly thanks to developments in path planning, obstacle avoidance and environment mapping.
However, the challenges will only increase as we use robots more and more to explore hostile places.
In the future, robots will be required to operate in environments that are not only undiscovered, but that the very nature of the environment is not well understood.
For example, robots in underground environments must cope with rough terrain, narrow confines and hampered light and communication. Robots in nuclear power plants will need to cope with high levels of radiation and restricted space. Meanwhile robots exploring other planets will need to deal with an environment which is unknown and unmapped, with the added challenge of long delays with communication.
In all of these scenarios, resilient navigation systems which are capable of sensing, mapping, understanding and reacting to its surrounds autonomously will be vital.
The exploration robots of the future will also need to be able to handle failures, and adapt its function to suit. Additionally, they’ll need to develop innate abilities to recognise new discoveries, and use that new knowledge to continue its task.
6. AI for robotics
Many of the challenges of the future will require the continued development of artificial intelligence.
Already robots are capable of object recognition using pattern recognition. However, while these fulfil one narrow aspect of intelligence, there is still a long way to go to replicate and exceed all the types of intelligence we see in humans and other animals.
To achieve this, scientists will need to create a comprehensive map of the key mechanisms of human intelligence, and recreate that into a software system.
7. Brain-computer interfaces
Brain-computer interfaces are a direct connection between brain and machine, allowing the machine to be controlled by direct thought.
Using BCI’s, we could in the future augment human ability, and, more importantly, restore function to patients with reduced capabilities. Restoring movement for paralysed people, controlling prostheses, and restoring neural function following strokes and other neurological injuries, are all occurring now in some form.
The challenge in the future however will be developing the technology to allow wider adoption Equipment for sensing of brain signals is currently expensive cumbersome, however work in happening into implantable microsensors with flexible electronics, ultra-low power needs, and wireless data.
Another challenge surrounds the processing of that data being collected. The folding and function of the brain’s cortex differs between people, making processing of signals an individually tailored requirement. Currently, long period of training, calibration and learning are required.
8. Social interaction
Social interaction is one of the most complex of human behaviours. However, it is one that because we’re so adept at recognising and interpreting social behaviour and norms, we underestimate its actual complexity.
There is so much nuance, rapid changes and unconscious social cues that we innately pick up, that it is argued our understanding of human social interaction is less understood than Newtonian mechanics or even human visual perception.
Social signals are also very context-dependent, and vary depending on cultures. Robots would not only need to pick up the social cues being displayed, but also adapt to cultural differences and learn an appropriate response to the setting.
Emotional responses, such as empathy, are another challenge for robots to understand and display as appropriate.
An added challenge is to be able to maintain solutions for these over a long-term period. At the moment most social robots are designed and programmed for short interactions with humans lasting minutes or hours at most. However, as robots become more integrated into society, the timescale of these interactions increases to become years. The social programming of a robot will need to expand from moment-to-moment engagements to being able to form and maintain long-term relationships.
9. Medical robotics
One of the fields where robotics will have its biggest impact will be medicine, improving healthcare and reducing cost.
One of the biggest challenges will be moving towards systems which show greater and greater levels of autonomy.
Where autonomous robots operate currently usually involves product manufacturing and similar roles which have been tailored to suit the robot’s capabilities. However, in medicine, the situation is always changing, uncontrolled and unique from patient to patient.
As a result, the robotics used today are usually focussed on enhancing the skills of a surgeon, where input is provided by a surgeon, which is carried out more precisely by the robot than is possible by human hand.
One potential for increased robotic use is having one surgeon supervise a set of robots who carried our certain tasks autonomously, but call upon the surgeon to take over during critical, patient-specific steps.
Moving away from surgery, implantable miniaturized multifunctional devices are being developed. These would be implanted long-term, monitoring and potentially intervening in medical episodes as they happen. However, these devices require long-term power availability, and also face the challenge of biocompatibility – being compatible with the body and not causing issues themselves. Additionally, they would need to detect and respond autonomously to all possible failures within the device itself.
Micro- or nano-robots which swarm through the patient’s body are another possibility being explored. However these robots would need to be biodegradable but non toxic, plus have the ability to target the diseased location and deliver a meaningful therapy.
10. Robot ethics and security
The potential regulatory, ethical and legal barriers facing the increased use of autonomous robots are some of the most important challenges facing the field.
In particular, there were 5 issues identified.
Firstly, excessive reliance on robotics could see sensitive tasks which should require a human guide at least supervising the task, being delegated entirely to AI.
Robotics and AI could see an increase in humans not taking responsibility for failures. Instead, responsibility is needed to be distributed in a different way.
Unemployment could well become a major factor in the roll out of increased robotics. A change in the workplace structure, a shift in skills requirements, and a potential deskilling of a workforce, are all possible in the near future. However, ethicists say this would be irresponsible, especially in safety-critical uses. Pilots should continue landing planes so that they are still able to should the AI be incapable, and radiologists should keep studying images to prevent deskilling in case the AI fails. Maintaining a skill base will remain critical.
Increased unemployment is probably inevitable however, with reports of 400 to 800 million jobs being at risk from AI. A shift in society expectations is critical in this scenario, where the financial benefits of this huge transformation are shared with those adversely affected, lowering inequality.
Fourth, AI is at risk of eroding human freedom. Rather than allowing humans to head in our own direction, human behaviour could need to change to accommodate automation.
Finally, one of the major risks is straight-out misuse. A very human problem, there is nevertheless the potential for AI to be misused in unethical ways. Some ethicists say that to minimise this chance, AI needs to be designed and used to treat every human being as an end, and never only as a means.
Robotics and AI is coming, and within our lifetimes will make a seismic shift to society. By tackling these 10 challenges, we will maximise the opportunities available to use, and hopefully, minimise the downsides to the coming robotic revolution. However, many of the problems we face are human, not technological, but likewise the solutions are also human, not technological. The future will be what we will make it.
Read more at Science Robotics