The NCE Focus Issue on Bioinspired Adaptive Intelligent Robots aims at collecting evidence of the different forms of biomimicry in robotics, from soft robotics and embodiment to neuromorphic sensing, computation and control, as enabling approaches to intelligent and adaptive robots.

The NCE Focus Issue on Bioinspired Adaptive Intelligent Robots is motivated by the increasing use of biomimicry in robotics to build adaptive and intelligent robots, able to safely move and interact with unknown environments. Bioinspiration can come at different levels, starting from the use of compliant and soft materials, the exploitation of morphology and embodiment to enable the execution of complex actions and ease sensing and perception, up to the use of biologically inspired sensors and processing, under the form of neuromorphic technology and computation. The scope of this collection of articles is to bring together most of the different forms of bioinspiration and promote their synergistic integration. It covers examples of bioinspired and neuromorphic pipelines for designing robots endowed with complex behaviour, extending also to discussing possible directions in medical applications, such as neurodegenerative diseases early detection.

An important trend in bioinspired robotics based on neuromorphic computing is to deploy full neuromorphic sensor-to-actuation pipelines, whereby event-driven spiking sensors feed asynchronous networks of spiking neurons that control the robot's motors. In such systems, latency, and efficiency are improved when the spiking neural networks are implemented on neuromorphic hardware. For those systems where the hardware implementation is not yet available, the networks are optimized for their future deployment.

In the first article, D'Angelo et al (2022) implement a neuromorphic pipeline exploiting the asynchronous output of the event-driven cameras to drive the visual attention of a humanoid robot. In an attempt to further decrease the system's latency, the pipeline is implemented on the SpiNNaker digital neuromorphic computing platform. Comparison to its implementation on GPU shows comparable attentional output, but with an average of 16 ms latency to produce a usable saliency map.

A fully neuromorphic closed-loop system is presented in Schoepe et al (2023) that is able to locate sound sources, such as human speech, by turning towards the direction of the sound source with a velocity linearly proportional to the angle difference between the sound source and binaural microphones. After this initial turn, the robotic platform remains in the direction of the sound source. The system is implemented on an FPGA, and it consumes around 1 W. Similarly, in Stroobants et al (2022) a neuromorphic closed-loop pipeline, implemented on the Loihi digital neuromorphic device, is used to estimate the pitch and roll angles of a quadrotor in highly dynamic movements from six-degree of freedom inertial measurement unit data. With only 150 neurons and limited training on a dataset obtained using a quadrotor in a real-world setup, the network shows competitive results as compared to state-of-the-art, non-neuromorphic attitude estimators.

Another key aspect of neuromorphic robotics is the implementation of bioinspired learning. Zhu et al propose the first developmental model of torsional eye movements driven by efficient coding without the need for external rewards or supervision (Zhu et al 2022). The model can be also applied to an event-based camera's active vision system, enabling it to overcome an orientation misalignment between the left and right cameras without any manual calibration. Zhaoqi et al (2022) propose a simplified and robust model for hippocampus-like place cells generation, based on the oscillatory interference model concept, to implement a bio-inspired simultaneous localization and mapping (SLAM) system for mobile robotics that can be used to achieve the localization and path-tracking functionalities of SLAM.

To conclude, two perspectives open the way to new robotics and medical applications that can take advantage of a neuromorphic approach for more realistic models and, hence, their deployment in real-world scenarios. A historical perspective (Szczecinski et al 2023) gives a gentle introduction into the neuroscientific principles that underlie models and inspire neuromorphic robot controllers for legged robots. The other perspective (Tolu et al 2023) discusses the current trend of implementing tools for the pressing challenge of early diagnosis of neurodegenerative diseases and how brain-body models can help.

We hope this collection provides new insights into how to build a new generation of adaptive intelligent robotics by combining bioinspired models and their implementation on neuromorphic substrates with compliant mechanical structures and promote the integration of all diverse bioinspired techniques into a cohesive approach.