Fig. 1. (Color online) The comparison between artificial neural systems composed of neuromorphic devices and biological neural systems. The multi-terminal floating-gate memristor in the novel structured device reproduced with permission^[43]. Copyright 2023, Springer Nature. Hardware Kalman filter circuit reproduced with permission^[44]. Copyright 2020, The American Association for the Advancement of Science. Biomimetic drive system reproduced with permission^[45]. Copyright 2022, The American Association for the Advancement of Science. The intelligent control system for the car reproduced with permission^[46]. Copyright 2023, The American Association for the Advancement of Science.

Fig. 2. (Color online) The schematic diagrams of different memristor switching mechanisms. The schematic diagrams shown in (a)−(d) are all based on the principle of ion migration memristors. (a) The structure of a memristor is based on electrochemical metallization principles. It involves the process of cation migration to form conductive filaments, which is closely related to conductivity. (b) The structure of a memristor is based on the principle of valence change. It differs from cation-based conductive filaments by using inert electrodes and facilitated conductive filament formation by anions. (c) The memristor's structure is based on interface transition of the non-filamentary type. (d) Illustration of electric double layer (EDL) charge accumulation at the interface between an ionic liquid and a semiconductor. By applying positive and negative gate voltages separately, cations and anions can accumulate electrostatically on the channel surface. Ionic gating memristor reproduced with permission^[50]. Copyright 2009, Springer Nature. (e) Diagram illustrating the switching process of MTJ in a spin memristor. Spin memristor reproduced with permission^[54]. Copyright 2022, Wiley-VCH. (f) Description of the switching process of FTJ in a ferroelectricity memristor. (g) An explication delineating the characteristics of phase-change memristor.

Fig. 3. (Color online) (a) Various 2-terminal and 3-terminal, multi-terminal devices serve as artificial synapses, including RRAM, Fe-FET, ion liquid-gated, etc. As well as crossbar arrays composed of memristors. (i) Ta/HfO₂ memristor reproduced with permission^[78]. Copyright 2021, Springer Nature. ZIF-8film memristor reproduced with permission^[68]. Copyright 2022, American Chemical Society. Bi₂Se₃ memristor reproduced with permission^[69]. Copyright 2023, Elsevier. (ii) Electret-based synaptic transistor reproduced with permission^[79]. Copyright 2020, American Chemical Society. Double-Gate memristor reproduced with permission^[80]. Copyright 2020, American Chemical Society. 3-TENG memristor reproduced with permission^[81]. Copyright 2023, Wiley-VCH. (iii) Multi-terminal Li_xMoS₂ memristor reproduced with permission^[56]. Copyright 2018, Wiley-VCH. Multi-gate artificial synaptic multiplexing unit reproduced with permission^[82]. Copyright 2022, The American Association for the Advancement of Science. Multi-terminal α-In₂Se₃ ferroelectric memristor reproduced with permission^[55]. Copyright 2021, Wiley-VCH. Coupled multiterminal oxide-based neuro-transistor reproduced with permission^[71]. Copyright 2019 Wiley-VCH. OASTs memristor reproduced with permission^[70]. Copyright 2023, Springer Nature. Multi-terminal MoS₂ synaptic transistor reproduced with permission^[72]. Copyright 2022, Wiley-VCH. (IV)The crossbar array comprised of 2 terminal devices reproduced with permission^[74]. Copyright 2019, Springer Nature. The crossbar array comprised of 3 terminal devices reproduced with permission^[75]. Copyright 2020, Science Press. (b) The various characteristics embodied by memristor devices include analog controllability with continuous adjustments, non-volatile behavior, synaptic properties, as well as (iii) spatiotemporal features derived from multi-modulation and multi-terminal inputs. (i) The illustration of conductance state response under potentiating pulses reproduced with permission^[44]. Copyright 2020, The American Association for the Advancement of Science. The graphical representation of the conductance levels after a series of pulsed voltage applied to the gate terminal reproduced with permission^[83]. Copyright 2021, The American Association for the Advancement of Science. The illustration of Retention tests of the device with multilevel conductance states reproduced with permission^[44]. Copyright 2020, The American Association for the Advancement of Science. (ii) Long-term plasticity behavior reproduced with permission^[68]. Copyright 2022, American Chemical Society^[76]. Copyright 2022, Elsevier. Devices’ multi-regulation characteristics reproduced with permission^[77]. Copyright 2023, Springer Nature^[76]. Copyright 2022, Elsevier. (iii) Characteristics of multi terminal devices reproduced with permission^[84]. Copyright 2023, Springer Nature^[72]. Copyright 2022, Wiley-VCH^[70]. Copyright 2023, Springer Nature^[71]. Copyright 2019 Wiley-VCH.

Fig. 4. (Color online) (a) (i) The inspiration for the connectionism of the Braitenberg vehicle, which mimics the foraging behavior of insects, (ii) training process of 2 x 2 memristor weights and the learning process of Braitenberg vehicle tracing rules, (iii) flowchart of the Braitenberg vehicle's supervised learning process along with detailed circuit diagrams incorporating the neuromorphic circuit processor and PWM drivers: The Bio-inspired processor receives signals from the left and right grayscale sensors of the Braitenberg vehicle. These signals pass through the connected 2 × 2 memristor processing module, before being inputted to the differential amplification circuit for computation. The resultant calculation is compared to a triangular carrier wave, generating PWM signals with varying duty cycles to control the steering of the servo motor. When an erroneous steering occurs, the supervisor provides feedback to modify the conductance values of the memristor array, thereby completing the training process for the tracing procedure. Revised illustration reproduced with permission^[67]. Copyright 2019, Wiley-VCH. (b) (i) Detailed schematic of the path-planning robot system, (ii) flowchart of the reinforcement learning process for the path-planning robot, (iii) training process of visual-motor association formed by the path-planning robot. Revised illustration reproduced with permission^[83]. Copyright 2021, The American Association for the Advancement of Science.

Fig. 5. (Color online) Biomimetic training effects based on synaptic plasticity. (a) The associative learning behavior is facilitated by synaptic plasticity based on memristors, thereby endowing robots with rapid response capabilities. (i) Revised illustration of ANP reproduced with permission^[68]. Copyright 2022, American Chemical Society. (ii) Revised illustration of ANP reproduced with permission^[76]. Copyright 2022, Elsevier. (b) The associative learning is facilitated by synaptic plasticity based on memristors, which endow robots with the ability to adjust sensitivity based on external environmental stimuli. (i) Process of gradually adapting to the environment reproduced with permission^[88]. Copyright 2022, Wiley-VCH. (ii) Training of associative learning reproduced with permission^[89]. Copyright 2022, The American Association for the Advancement of Science. Involuntary reflex protective mechanism under harmful stimuli reproduced with permission^[69]. Copyright 2023, Elsevier.

Fig. 6. (Color online) Memristor's multi-modulation mechanisms and device structure endow robots with the brain-like capability of multi-information integration. (a) (i) Based on the varied gating capabilities arising from memrisors’ multi-modulation mechanisms reproduced with permission^[76]. Copyright 2022, Elsevier. (ii) Based on the memristors’ structure reproduced with permission^[84]. Copyright 2023, Springer Nature. (b) Implementing spatiotemporal information integration in memristor-based neural-mimicking circuits based on peripheral circuitry and device structure to achieve the distinction of sound azimuth angles. (i) Based on the peripheral circuit reproduced with permission^[100]. Copyright 2018, American Association for the Advancement of Science. (ii) Based on memristors’ structure reproduced with permission^[70]. Copyright 2019 Wiley-VCH.

Fig. 7. (Color online) Parallel computing achieved through memristor array structures and novel device architectures. (a) Parallel processing of batches of images is achieved through a memristor crossbar array structure coupled with frequency-modulated carriers reproduced with permission^[78]. Copyright 2019 Wiley-VCH. (b) By leveraging device structural characteristics, achieve the functionality of controlling multiple drives in parallel with a single input signal. Revised illustration reproduced with permission^[71]. Copyright 2023, Springer Nature.

Fig. 8. (Color online) Biomimetic control system based on memristors. (a) The biomimetic local analog control system includes (i) oculomotor nerves reproduced with permission^[103]. Copyright 2023, The American Association for the Advancement of Science. (ii) A pair of working muscles reproduced with permission^[45]. Copyright 2022, The American Association for the Advancement of Science. (iii) A closed-loop control system reproduced with permission^[82]. Copyright 2022, The American Association for the Advancement of Science. (b) A holistic neuromorphic mixed-signal control system platform is inspired by the cerebrum and cerebellum, implementing sensor data fusion and control algorithm adjustments. Revised illustration reproduced with permission^[44]. Copyright 2020, The American Association for the Advancement of Science^[104]. Copyright 2022, Springer Nature^[105]. Copyright 2023, RSC Pub^[106]. Copyright 2023, Elsevier.