This study presents a sensorimotor control neural circuit (SCNC) inspired by the avoidance behavior of crickets, using three memristive Hodgkin-Huxley (H-H) neurons to enable robotic obstacle avoidance. The SCNC consists of an ascending neuron with mixed spiking and bursting features, a bursting-detection neuron (BDN), and a spiking-detection neuron (SDN). The ascending neuron encodes sensory input, while the BDN and SDN decode the mixed firing patterns to control the robot's steering and actuating motors. The SCNC achieves a latency reduction of more than one order of magnitude compared to conventional platforms, demonstrating the efficiency of emulating neural circuits for intelligent tasks. The SCNC is implemented using NbO₂ memristors, which exhibit rich dynamics, low power consumption, and good scalability, making them suitable for constructing H-H neurons with high bio-plausibility. The H-H neuron circuit comprises two resistors, two capacitors, and two NbO₂-based threshold switching (TS) memristors. The TS devices simulate the sodium and potassium ion channels in biological neurons. The circuit parameters are carefully designed to enable the neuron to exhibit both spiking and bursting features, with the bursting ratio increasing as the input intensity increases. The SCNC is integrated into a robot to perform obstacle avoidance, with the BDN and SDN firing frequencies controlling the steering and actuating motors, respectively. The closer the obstacle is to the robot, the larger the steering angle and the slower the speed. The SCNC successfully emulates the mixed firing patterns of AN₂ neurons, demonstrating the potential for encoding emergencies and achieving real brain-like systems. The SCNC also incorporates a selective communication scheme, where the BDN and SDN decode the mixed firing patterns from the ascending neuron. The BDN responds to bursting features, while the SDN responds to spiking features, enabling the robot to react to different obstacle distances. The SCNC achieves a latency reduction of more than 50 times compared to conventional methods, with a minimum delay of 100 μs. The power consumption of the SCNC is significantly lower than that of FPGA-based systems, making it a promising solution for mobile intelligent robots. The study highlights the potential of memristive neurons for implementing real brain-like systems and paves the way for next-generation intelligent machines. The SCNC demonstrates the feasibility of using firing feature-driven neural circuits for robotic obstacle avoidance, with the potential for further improvements in scalability, plasticity, and energy efficiency.This study presents a sensorimotor control neural circuit (SCNC) inspired by the avoidance behavior of crickets, using three memristive Hodgkin-Huxley (H-H) neurons to enable robotic obstacle avoidance. The SCNC consists of an ascending neuron with mixed spiking and bursting features, a bursting-detection neuron (BDN), and a spiking-detection neuron (SDN). The ascending neuron encodes sensory input, while the BDN and SDN decode the mixed firing patterns to control the robot's steering and actuating motors. The SCNC achieves a latency reduction of more than one order of magnitude compared to conventional platforms, demonstrating the efficiency of emulating neural circuits for intelligent tasks. The SCNC is implemented using NbO₂ memristors, which exhibit rich dynamics, low power consumption, and good scalability, making them suitable for constructing H-H neurons with high bio-plausibility. The H-H neuron circuit comprises two resistors, two capacitors, and two NbO₂-based threshold switching (TS) memristors. The TS devices simulate the sodium and potassium ion channels in biological neurons. The circuit parameters are carefully designed to enable the neuron to exhibit both spiking and bursting features, with the bursting ratio increasing as the input intensity increases. The SCNC is integrated into a robot to perform obstacle avoidance, with the BDN and SDN firing frequencies controlling the steering and actuating motors, respectively. The closer the obstacle is to the robot, the larger the steering angle and the slower the speed. The SCNC successfully emulates the mixed firing patterns of AN₂ neurons, demonstrating the potential for encoding emergencies and achieving real brain-like systems. The SCNC also incorporates a selective communication scheme, where the BDN and SDN decode the mixed firing patterns from the ascending neuron. The BDN responds to bursting features, while the SDN responds to spiking features, enabling the robot to react to different obstacle distances. The SCNC achieves a latency reduction of more than 50 times compared to conventional methods, with a minimum delay of 100 μs. The power consumption of the SCNC is significantly lower than that of FPGA-based systems, making it a promising solution for mobile intelligent robots. The study highlights the potential of memristive neurons for implementing real brain-like systems and paves the way for next-generation intelligent machines. The SCNC demonstrates the feasibility of using firing feature-driven neural circuits for robotic obstacle avoidance, with the potential for further improvements in scalability, plasticity, and energy efficiency.