Using robots to locate odor sources or smoke sources is considered a safe alternative to human and animal rescuers in high-risk search and rescue missions.
Compared to static sensor networks, mobile robots can significantly reduce the number of sensing nodes and improve the accuracy of location estimation.
In addition to developing dedicated robotic platforms, researchers have invested considerable effort in the development of odor source localization algorithms. Since the early 1990s, various odor source localization algorithms have emerged, including concentration gradient climbing algorithms, bio-inspired algorithms, probabilistic methods, and algorithms suitable for multi-robot systems.
However, the analog signals collected by sensors need to be converted into digital signals by an analog-to-digital converter, which are then stored and transmitted to the processing unit.
Between the sensing nodes and the external processing unit, the large amount of perceptual data exchanged under limited bandwidth severely restricts processing speed, leading to high energy consumption and potential safety hazards, which is in stark contrast to the efficient perceptual processing methods in biological systems.
One of the most critical environmental variables in locating odor sources or smoke sources is wind speed. Mimicking the wind-seeking behavior exhibited by rodents during odor source searches is a method for intelligent robots to track efficiently.
In natural environments, animals actively manipulate their tactile organs to convert tactile information into electrical signals. Mice explore their environment with their whiskers, sweeping approximately 5 to 10 times per second to perceive, locate, and identify objects.
During movement, each whisker interacts dynamically with the external environment, which can be sensed by thousands of mechanoreceptors distributed around the hair follicle-sinus complex.
These receptors transmit signals to 150 to 400 neurons located in the trigeminal ganglion (NV), forming the primary processing stage of the whisker perception system.
The mechanical response of mouse whiskers to airflow reveals information about wind direction and speed.
When air flows over the whiskers, it causes them to bend in the direction of the wind, with the degree of bending related to wind speed, after which the whiskers oscillate from that deflected position. The whisker system of terrestrial mammals has greatly inspired the design of intelligent wind-seeking systems.
Mimicking the wind perception mechanism of mice has many advantages. First, the whiskers on the mouse’s face are arranged in a regular row distribution, forming a structured array.
Each whisker is connected to thousands of mechanoreceptors that can convert tactile stimuli into electrical signals and transmit them to the brain via the primary afferent nerves of NV. In NV, each neuron responding to whisker movement corresponds to only one whisker.
The neural responses are orderly transmitted to the primary somatosensory cortex of the brain. Secondly, the whisker system is very suitable for studying cellular responses in natural exploratory behavior, which is the ultimate goal of related research.
This system makes tracking and recognizing perceptual inputs more precise. Finally, compared to the primary afferent nerves entering the spinal cord of primates, the NV of mice is located at the base of the brain, making it easier to study and obtain.
To construct an artificial whisker system with wind-seeking behavior, it is necessary to combine wind sensors with artificial NV neurons (NVNs).Dr. Han Suting’s team from the Department of Applied Biology and Chemical Technology at The Hong Kong Polytechnic University introduces a pulse-based wind direction tracking system consisting of two core components: self-powered carbon black (CB) sensors and memristors with threshold switching characteristics (HfO₂).
Combined with self-powered sensors, these types of neuronal devices can simultaneously sense wind speed, humidity, and temperature signals, encoding them in the form of pulses at different frequencies.
Finally, the authors mounted this artificial whisker system on a robotic vehicle, demonstrating its strong tendency for straight-line movement when approaching a wind source. Additionally, the angular variation between the robotic vehicle and the wind source shows that even starting from a random initial state, it can accurately perceive wind direction. The research results indicate that the artificial whisker system possesses wind-seeking behavior capabilities, and the hardware-implemented NV neurons significantly enhance this capability.