Multiple drones or mobile robots acting in concert are opening up new applications for automation, covering larger areas, delivering more useful sensor data, and interacting with their environments in new ways. In the U.S., the military has been a leader in developing swarm robots, but academia also plays a significant role.
One of the key differences is that many swarm robotics developments within academia have been around the software “back end.” The U.S. military, on the other hand, has worked on both algorithms and hardware.
For instance, Georgia Tech has developed algorithms to power swarm robots. One of them is focused on ensuring that such devices can find a balance between safety and performance. Small autonomous robots used it to move and “switching places” without bumping into each other or creating blocks. Researchers also saw that even if one robot acted “wild,” the rest of the pack would adapt and change their movements accordingly.
Georgia Tech is also behind swarm robots that can play the piano. Unveiled back in 2012, Khepera III was programmed to be at the most important location in the shortest amount of time, to ensure that it would in the right place to hit the correct keys. This seems simple, but such decision-making is a precursor of predictive intelligence.
Cost and complexity of swarm robots
In 2014, the University of Colorado Boulder unveiled a plan for swarm robots called “Droplets” and raised money via crowdfunding. These robots were self-charging — by moving their legs, they charged their hardware. They could work in groups of up to 100 or 1,000.
The problem, however, was the cost and infrastructure required, such as the powered floor, infrared communications, overhead camera, and projector system. For just 100 Droplets, the cost was an estimated $10,000.
Two years later, Stanford unveiled “ant robots” that worked as a team. Six of them worked together to pull a 1.8-ton car, with a driver on board, even though each ant robot weighed only 17 grams, meaning each robot pulled a mass 270 times its own weight.
Stanford isn’t the first American university to base robots on ants. In 2015, Princeton University conducted research into the complexity of ant behavior, such as how they use their bodies to construct objects like bridges. The researchers proposed that further understanding of ants can help in creating “intuitive robots that can cooperate as a group.”
Decentralized planning and blockchain
In 2016, MIT unveiled a new planning algorithm that groups of robots could use to improve efficiency in their operations. Unlike other planning algorithms that are centralized or rely on a single computer to handle delegating tasks, MIT’s algorithm is decentralized, i.e., each robot makes its own decision.
Decentralized algorithms are harder to develop because it’s harder to control what decisions each robot will make, but they can enable robots to less bandwidth when communicating. They also allow them to remain online in situations where power or communications have been severed — like in busy factory or on a battlefield.
While some researchers focus on developing robot intelligence, others are looking to tap emerging technologies like blockchain. In simple terms, blockchain is a highly secure and public “ledger” that records information like transactions. Blockchain has mainly been viewed as a new way to improve security and transparency around the use of cryptocurrencies like Bitcoin.
A white paper published by an MIT research affiliate described blockchain as a middleman to help coordinate swarm robots as they complete tasks. One way to integrate blockchain is to make blockchain responsible for delegating tasks if no robot has been given a “leadership role. Delivery drones, marine vessels, and self-driving vehicles are possible applications for this technology.
More recently, researchers at the University of Buffalo, Harvard University, and the University of Florida are working to improve RoboBees’ perception and navigation. The project is funded with a $1.1 million grant from the National Science Foundation and promises to pool individual robots’ intelligence.
Mind control and open ideas
Arizona State University has taken swarm robotics to the next level by integrating another equally revolutionary technology: mind control. ASU has developed a skull cap that enables a human thought about movement to be converted into a signal that can be sent to a drone. Initially, one person could command of four drones. In the future, will it be dozens?
When it comes to swarm robotics in U.S. academia, there are several things to consider. First, academia is most interested in pure scientific research and less in commercialization or military uses.
Second, academia, unlike the government, is restricted when it comes to funding and approval for ideas. It has to work with partners across institutions, disciplines, and industries.
Third, there has historically been little information around how innovations such as swarm robots developed by academia will integrate with business or government markets. This poses challenges, but also opportunities. Research teams have become increasingly aware of the need to identify and collaborate with prospective end users.
For example, Oregon State University is part of a $7.1 million Defense Advanced Research Projects Agency initiative to develop swarm robots to support troops in the field.
In the case of the U.S. military, swarm robotics could help it maintain its strategic edge. For businesses, swarm robots are a building block for the Industrial Internet of Things, predictive analysis, and ultimately, more efficient operations.