Oscillatory Neural Networks for Obstacle Avoidance on Mobile Surveillance Robot E4

Abstract

HAL is a multi-disciplinary open access archive for the deposit and dissemination of scientific research documents, whether they are published or not. The documents may come from teaching and research institutions in France or abroad, or from public or private research centers. L'archive ouverte pluridisciplinaire HAL, est destinée au dépôt et à la diffusion de documents scientifiques de niveau recherche, publiés ou non, émanant des établissements d'enseignement et de recherche français ou étrangers, des laboratoires publics ou privés.

Oscillatory neural network learning for pattern recognition: an on-chip learning perspective and implementation

“…Furthermore, the next developments will focus on possible applications with the ONN on-chip learning architecture. The digital ONN design has already been used for sensor data treatment in various applications, like interfacing with a camera for image recognition (Abernot et al, 2021 ) or using proximity sensor data to perform obstacle avoidance (Abernot et al, 2022a ), so we are confident on the integration of our architecture with different sensors. Possible applications for the digital ONN on-chip learning architecture could be in the robotics domain where real-time continual learning is often necessary, and where the digital ONN design already showcased good performances (Abernot et al, 2022a ).…”

“…The digital ONN design has already been used for sensor data treatment in various applications, like interfacing with a camera for image recognition (Abernot et al, 2021 ) or using proximity sensor data to perform obstacle avoidance (Abernot et al, 2022a ), so we are confident on the integration of our architecture with different sensors. Possible applications for the digital ONN on-chip learning architecture could be in the robotics domain where real-time continual learning is often necessary, and where the digital ONN design already showcased good performances (Abernot et al, 2022a ). For example, navigation, in the context of mobile robots, is a complex task depending on the environment, where continuous learning is necessary to adapt to evolving situations.…”

“…For example, navigation, in the context of mobile robots, is a complex task depending on the environment, where continuous learning is necessary to adapt to evolving situations. A first proof of concept of two pre-trained cascaded ONNs performing obstacle avoidance from proximity sensors was shown in Abernot et al ( 2022a ). Though in Abernot et al ( 2022a ), the pre-trained ONNs are capable of finding a novel direction using information from 15 proximity sensors whose configurations are used to define the training patterns.…”

“…In particular, a digital ONN implementation on FPGA was introduced in Abernot et al ( 2021 ) to explore novel ONN architectures, learning algorithms, and applications. The digital ONN on FPGA showcased fast and efficient computation for edge applications, for example performing obstacle avoidance on mobile robots by reading proximity sensor information (Abernot et al, 2022a ), replacing convolution filters for image edge detection (Abernot and Todri-Sanial, 2023 ), or even accelerating the SIFT feature detection algorithm (Abernot et al, 2023a ). We believe ONN implementation on FPGA is attractive for real-time applications for which providing on-chip learning is important.…”

Enhancement of temperature-modulated NbO2-based relaxation oscillator via interfacial and bulk treatments

“…CMOS oscillator devices such as ring oscillators, though well established, are facing problems such as high power density and the lack of frequency tunability [2]. There are several applications for the ONN, such as an oscillatory neural autoencoder [10], solving cluster analysis problems [11], robotic controls [12,13], and pattern recognition [14][15][16]. The frequency tunability of a device is essential in ONN applications particularly in pattern recognition based on the frequency shift keying scheme.…”