Giant, Six-legged Robotic Fly Developed To Make AI Learning Quicker

Engineers are using a giant six-legged robotic fly to develop quicker AI-powered machines.

The bug-like robot mimics fruit flies – the most commonly studied insect in biology – in a bid to improve machine learning.

Scientists hope to shed light on how insects sense forces in their limbs while walking in order to help create more mobile robots.

The Drosophibot prototype opens the door to developing quicker-moving AI (artificial intelligence) machines for industry, researchers said.

It may also lead to more agile drones for a host of uses – ranging from quickly surveying a disaster area to delivering packages more efficiently to customers.

When humans run, their legs exhibit minimal contact with the ground. Six-legged insects, however, run fastest when using a three-legged, or ‘tripod’ gait.

Force receptors in their limbs known as campaniform sensilla (CS) respond to stress and strain – providing important information for controlling locomotion.

In humans and other mammals organs in the tendon do a similar job – relaying information about force levels to the central nervous system.

Lead author, Dr. Nicholas Szczecinski, of West Virginia University, said: “I study the role of force sensors in walking insects because these sensors are critical for successful locomotion.

“The feedback they provide is critical for proper posture and coordination.”

None airborne flies have a tripod gait with three legs on the ground at all times – two on one side of their body and one on the other.

Engineers have long believed copying it would make six-legged robots move fastest.

Robots can model friction between moving parts and the inclusion of delays to send neural signals better than computer simulations.

The limbs also have the advantage of being able to record the sending and receiving of every single signal and resulting mechanical actions, which is not possible with animals.

Dr. Szczecinski added, “Walking is an inherently mechanical task, so understanding the neural control of walking requires simultaneously investigating mechanics and neural control.

“Properly functioning walking robots can serve as prototypes for machines that could help people farm in extreme terrains, explore other planets, or walk through forests to monitor their health.”

Drosophibot has anatomical aspects not present in other, similar bio-robots including a retractable abdominal segment.

It also has insect-like dynamic scaling and compliant feet segments to capture a complete picture of how campaniform sensilla monitors forces while walking.

The US team also uses a single robotic leg which allows for a simplified simulation of the sensory experience while walking.

Dr. Szczecinski also explores the role of CS in real insects by isolating their limbs and monitoring sensory pathways with electrodes when different forces are applied.

These recorded sensory signals are then used to develop models for the robotic legs.

Dr. Szczecinski said, “By recording their response to many different signals, we can paint a clearer picture of how they convert forces into neural activity.

“We use many different stimuli because the CS are highly dynamic and are always adapting to the applied forces.”

Dr. Szczecinski’s research has revealed very strong similarities between their real insects and robotic counterparts.

He said, “We find that for every insect species we check, our model is equally well-equipped to describe the way the CS turns forces into neural activity.

“This suggests that each species’ organs are broadly functioning in the same way.”

Drosophibot was presented at a meeting of the Society for Experimental Biology in Edinburgh.

Produced in association with SWNS Talker

Edited by Kyana Jeanin Rubinfeld and Jessi Rexroad Shull