Mitigating Neural Habituation in Insect Bio-Bots: A Dual-Timescale Adaptive Control Approach

Area of Science:

• Bio-cybernetics
• Robotics
• Artificial Intelligence

Background:

• Bio-cybernetic organisms integrate biological movement with electronic control.
• Challenges include individual variability and stimulus habituation.
• Existing systems lack dynamic adaptation to individual specimens.

Purpose of the Study:

• To introduce an Adaptive Neuro-Fuzzy Inference System (ANFIS) for bio-cybernetic applications.
• To dynamically calibrate control systems to individual arthropod specimens.
• To establish benchmarks for adaptive control in bio-hybrid platforms.

Main Methods:

• Developed and implemented an ANFIS using a miniaturized neural controller.
• Compared ANFIS performance against natural behavior and non-adaptive controls in *Gromphadorhina portentosa*.
• Assessed obstacle navigation, target acquisition, stimulus responsiveness, and biocompatibility over 14 days.

Main Results:

• ANFIS achieved 81% obstacle navigation efficiency, significantly higher than non-adaptive methods (42%).
• Effective behavioral modulation was sustained for 47 minutes, compared to 26 minutes for non-adaptive controls.
• Stimulus responsiveness was maintained 3.5-fold longer, with 73% target acquisition in complex terrain.

Conclusions:

• Adaptive neuro-fuzzy architectures significantly outperform conventional methods in bio-cybernetics.
• The ANFIS system demonstrates robust control and sustained responsiveness.
• This research enables advanced bio-hybrid platforms for confined-space search-and-rescue operations.