There’s power in numbers. It’s a fact that’s true for humans and more so for other species that thrive when assembled in vast numbers like ants, birds and fish. Such swarms rely on the group’s collective intelligence to survive. It is a concept that has caught on in the world of robotics, where researchers are devising robotic swarms to operate effectively in places humans are reluctant to venture.
The field of robotics often conjures images of mechanical moving arms in manufacturing, food delivery bots or the much more advanced humanoid counterpart of robots that still primarily reside in the realm of science fiction. Yet, there’s also the need for a class of small, nimble and unassuming bots, which are effective at search-and-rescue operations in hazardous environments, extraterrestrial exploration, subterranean inspection and even agricultural tasks like planting and harvesting.
Yasemin Ozkan-Aydin leads a team of researchers at the University of Notre Dame that is working to overcome the limitations of ground-based robot swarms due to the particular challenges of terrestrial locomotion. A recent study demonstrated how robots that mimic multi-legged organisms—configured with directionally flexible legs and tails and linked together autonomously—can navigate variable terrain without complex control or sensing.
“Understanding the emergent capabilities in both living and artificial collectives remains a frontier challenge in science and robotics,” said Ozkan-Aydin. “While numerous swarm robots have been built that can move and perform tasks in water or air, there has been limited research into the use of robots to swarm in complex terrestrial environments. The state of the art has dealt mainly with wheeled robots in flat laboratory environments.”
This terrestrial clumsiness of robots may seem surprising since humans are adept at locomotion. However, considerable advancement has been made in the arena of aerial and aquatic swarm robots, as air and water are more homogenous environments. For example, a Harvard University research team recently unleashed a school of silicon robotic 3D-printed fish that can swim together in coordination and observe wildlife. Meanwhile, flying robots are already busy in the field equipped with heat sensors to detect wildfires.
Yet, terrestrial environments can be complex and unpredictable, especially on rough terrain outside the built environment. For this reason, the Notre Dame researchers focused on the biomechanics of animal locomotion. They were inspired by multi-legged organisms that can cooperate to solve problems like traversing obstacles and maneuvering through tight spaces.
“The ultimate aim of this study is to use the robots in realistic environments including rough terrain and subterranean, where the robots navigate challenging and previously unseen, dynamic environments,” explained Ozkan-Aydin.
Although prior research on terrestrial swarm robots has been conducted on smooth terrain, Ozkan-Aydin and her team demonstrated that a swarm of chain-able legged robots can walk on more challenging, real-world surfaces and complete tasks that would be too challenging for solitary bots.
The insect-like mechanical design of the bots with flexible tails and legs allows for movement proficiency in simple environments even without the use of sensors. In addition, the tail provides air-righting capabilities, greater stability on uneven surfaces and the ability to climb on walls. However, once the complexity of the terrain increases, the solitary bots falter. Conversely, grouping the robots into a swarm where they are connected allows the group to navigate rough terrain, traversing gaps, climbing stairs and transporting objects. In other words, when the tasks are simple, each individual bot can operate independently, but if the situation becomes more challenging, the individual bot units organize into a larger, more effective multi-legged system.
In the demonstrations, each robot was equipped with two touch sensors and one light sensor. An Arduino-based controller, a battery and passive magnetic connectors enabled the bots to merge. Additionally, an open-loop controller, coupled with the bots’ individual structures, allowed them to walk in mostly smooth environments without sensory feedback. Each robot walked with a diagonal gait, so that its opposite side legs contacted the ground simultaneously. Each was also equipped with a passive tail for stability.
With the gap-traversal portion of the experiments, the solitary 22-cm-long robots struggled to traverse a 5-cm gap. However, chaining the bots together shifted the center of mass enough so they could travel across the gap. In terms of stairs, a single robot could climb a 1.25-cm-high stair but then got a rear leg or tail stuck on a 2.5-cm-high stair. However, a chain of three robots was able to climb the higher stair.
The researchers ran a single robot over scattered wooden blocks attached to a flat board to demonstrate its rough-terrain locomotion. Unfortunately, it repeatedly failed due to becoming stuck, lacking sufficient thrust to free itself. Finally, however, just two chained robots were able to navigate the course successfully.
One of the most significant advantages of the swarm robot model is teamwork. When robots operate independently in the field, their ability to rescue each other from getting stuck or losing a leg is crucial to the success of a mission. The researchers found that even if a leg of one of the bots was disabled, three chained robots could successfully operate despite the handicap caused by the individual robot. For example, in a search and rescue demonstration, one bot was designated a “searcher,” while the others were designated “helpers.” The searcher began walking toward a staircase in the direction of a light source. Once it became stuck on the stairs, it backtracked and sent a signal to a helper bot. The helper bot approached, and the two were able to connect and successfully climb the stairs together.
One of the biggest limitations of the technology is that the robots have a minimal communication range through light sensors. However, Ozkan-Aydin anticipates that as the field expands, the ability of the swarm bots to coordinate activities will advance, and they would be able to swap the roles of searcher and helper as conditions arise. Currently, if a helper robot falls outside of the beam of a searcher robot, the helper robot is unable to locate it and provide aid.
“In the future design, the communication between robots should be enhanced using other types of sensors such as GPS,” Ozkan-Aydin said. “However, as the complexity of the system increases, the robots become more difficult to control.”
Ozkan-Aydin emphasized that mechanical intelligence, defined as a mechanism that responds to the environment, adapts to new external situations, or performs some function without any sensor feedback from a controller, plays a vital role in the effectiveness of swarm robotics.
Powering the swarm is another area that could benefit from an upgrade for future design, Ozkan-Aydin said. The current setup uses a lithium polymer battery attached to the robot’s body joint. One possibility for future design includes integrating piezoelectric materials into the robots’ legs so that they can harvest energy as they walk. Solar panels atop the bots could also be used to recharge batteries in the field. A third option is to have only one robot equipped with energy harvesting equipment to transit power to other bots wirelessly.
Despite the need for future design upgrades and further studies, Ozkan-Aydin said that the use of low-cost robots and experimental setups in the current research could inspire others who have limited laboratory resources to make progress in the field of swarm robotics.