13 August 2020 in IEEE Signal Processing Letters
We address the problem of 3-D reconstruction using neuromorphic cameras (also known as event-driven cameras), which are a new class of vision-inspired imaging devices. Neuromorphic cameras are becoming increasingly popular for solving image processing and computer vision problems as they have significantly lower data rates than conventional frame-based cameras.We develop a neuromorphic-camera-based Fringe Projection Profilometry (FPP) system. We use the Dynamic Vision Sensor (DVS) in the DAVIS346 neuromorphic camera for acquiring measurements. Neuromorphic FPP is faster than a single-line-scanning method. Also, unlike frame-based FPP, the efficacy of the proposed method is not limited by the background while acquiring measurements. The working principle of the DVS also allows one to efficiently handle shadows thereby preventing ambiguities during 2-D phase unwrapping.
28 June 2024 in IEEE Robotics & Automation Magazine
Applications in robotics or other size-, weight-, and power-constrained (SWaP) autonomous systems at the edge often require real-time and low-energy solutions to large optimization problems. Event-based and memory-integrated neuromorphic architectures promise to solve such optimization problems with superior energy efficiency and performance compared to conventional von Neumann architectures. Here, we present a method to solve convex continuous optimization problems with quadratic cost functions and linear constraints on Intel’s scalable neuromorphic research chip Loihi 2. When applied to model predictive control (MPC) problems for the quadruped robotic platform ANYmal, this method achieves more than two orders of magnitude reduction in the combined energy-delay product (EDP) compared to the state-of-the-art solver, OSQP, on (edge) CPUs and GPUs with solution times under 10 ms for various problem sizes. These results demonstrate the benefit of non-von Neumann architectures for robotic control applications.
1 October 2017 in Journal of Computational and Theoretical Nanoscience
Increasing population and decreasing sources of conventional energy sources has resulted in decreased availability of energy per person. In this context, the role and relevance of solar energy has been increasing at a very rapid pace in the recent times. It is common knowledge that the output of a solar panel increases with irradiation. However increase in irradiation with ambient temperature increase also increases module temperature. In order to operate the solar panel at optimum efficiency, it is vital to monitor the rise in module temperature. This is because after a certain temperature, solar cells operate at decreased efficiencies due to negative thermal coefficient. This paper aims to compare and develop temperature prediction schemes for solar panels in the Indian environment and thus bring the error in prediction down in the range of 5–10% or less. The logged data for every second for a period of 12 days from 1st December 2013 to 12th December 2013, recorded at Adugodi, Bangalore, India is used as the primary reference for model development. A collection of mathematical models made by various researchers is tested for daily averages of data available for two different types of solar cells to test generality. Few well behaving models are shortlisted and tested for per second and five-minute average data. A workflow is developed and described for the same. The workflow gives a visual representation of the error, enabling the user to combine different models based on their regions of optimum operation. The concept of Artificial Neural Networks is applied and tested for its validity with the data in hand. The different schemes of prediction are then compared to pick the best one. The prediction schemes outlined in this paper give minimal error which is in the range of 3.5–5% for Neural Networks and less than 10% on an average for mathematical models.