What’s the secret to unlocking artificial intelligence (AI) and making it ubiquitous in our everyday lives? The answer may lie with the most abundant animals on earth — insects.
The behavioral adaptations of insects could help commercial organizations overcome a significant hurdle for AI adoption today: cost. The cost to design, produce, and implement this technology is still prohibitive for practical uses. As a result, research-focused competitions have played an important role in addressing current business challenges by applying new design techniques and strategies and showcasing new technology.
At the Silk Road Robotic Innovation Competition (SRRIC) held last month at Xi’an Jiao Tong University in China, a team of engineering students from Washington University in St. Louis demonstrated FlowBot, a low-cost, autonomous vehicle that avoids objects and navigates just like an insect does.
Insects lack the binocular vision of humans to perceive depth. Instead, they use an optical flow method to perform basic functions, such as obstacle avoidance, stable maneuvering, and navigation. While the insect brain is clearly not as sophisticated as the human brain, insects are among the most agile creatures on this planet. Inspired by this, a team of engineers sought to build a robust, low-power autonomous robot modeled after an insect’s nimble characteristics. One goal for the team of five students was to create a hardware system for intelligent robots at a much lower cost than current hardware platforms.
While there is an abundance of AI software startups, the design, assembly, and fabrication of hardware components for AI, in addition to the power consumption of hardware systems, are far from cheap. For the small and medium-size businesses that can’t afford to use the expensive, heavy-duty, and power-hungry sensors and processors used by current models of self-driving cars, for example, research into alternative technology platforms is critical to widespread adoption.
Universities from around the world gathered at this year’s SRRIC to demonstrate everything from smart farm robots and an amphibious vehicle for underwater operation to intelligent medical rehabilitation assistants and fabric-weaving robotics for large-scale material manufacturing. When some of the brightest engineering minds in the world are incentivized to further technology development and deepen international engagement, they discover synergies between various areas of expertise and identify opportunities to share resources across borders.
This competition, and many like it, demonstrates that making AI technology practical and useful for robotics and various IoT applications is not about building a car that is 100 percent reliable or a robot that never bumps into things, but rather about how to develop a system that is resilient and self-healing.
Some of the most successful examples of resilient robotic technology originated in nature. Raytheon, a defense technology company, has been testing autonomous robots with distributed brain structures designed after octopus brains to help the machines better adapt to their surroundings. This technology may one day help control autonomous drone swarms for critical missions.
Similarly inspired by how an insect perceives and navigates the world, the FlowBot runs an optical flow algorithm. FlowBot uses only a single front-facing camera to achieve autonomous obstacle avoidance capability on a Pi Car four-wheel robot. The efficient visual algorithm doesn’t require significant computational power — perfect for the Pi Car platform, which targets autonomy for small-scale systems.
This isn’t the first time insects have been used to advance robotic technology. Biological sensing systems are far more complex than their engineered counterparts, and engineers have leveraged the locust’s sense of smell to create new biorobotic sensing systems that could be used in homeland security.
Engineers aiming to create AI that’s as lifelike as possible should ask themselves: Why reinvent the wheel? By taking advantage of the biological solutions, engineers can cut down on costs for research and development, and align technology with processes that already exist in nature.
The simplest creatures, like insects, can offer insights and inspirations that significantly advance man-made systems in responsiveness, robustness, and energy efficiency. FlowBot and Pi Car have demonstrated tremendous opportunity for the development and production of cheap and easily reproducible AI swarm products that could bring AI into our daily lives to accomplish sophisticated tasks, even if not every single one succeeds.
Xuan “Silvia” Zhang is an assistant professor of electrical and systems engineering at Washington University in St. Louis.